WO2016111724A1 - Mémoire vive résistive avec canaux implantés et rayonnés - Google Patents

Mémoire vive résistive avec canaux implantés et rayonnés Download PDF

Info

Publication number
WO2016111724A1
WO2016111724A1 PCT/US2015/045731 US2015045731W WO2016111724A1 WO 2016111724 A1 WO2016111724 A1 WO 2016111724A1 US 2015045731 W US2015045731 W US 2015045731W WO 2016111724 A1 WO2016111724 A1 WO 2016111724A1
Authority
WO
WIPO (PCT)
Prior art keywords
switching
memory device
electrode
ions
switching region
Prior art date
Application number
PCT/US2015/045731
Other languages
English (en)
Inventor
Shih-Yuan Wang
Shih-Ping Wang
Original Assignee
Shih-Yuan Wang
Shih-Ping Wang
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shih-Yuan Wang, Shih-Ping Wang filed Critical Shih-Yuan Wang
Priority to CN201580027848.6A priority Critical patent/CN106463340A/zh
Priority to JP2016569376A priority patent/JP2018516447A/ja
Priority to EP15877278.0A priority patent/EP3243211B1/fr
Publication of WO2016111724A1 publication Critical patent/WO2016111724A1/fr

Links

Classifications

    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10B—ELECTRONIC MEMORY DEVICES
    • H10B63/00—Resistance change memory devices, e.g. resistive RAM [ReRAM] devices
    • H10B63/80—Arrangements comprising multiple bistable or multi-stable switching components of the same type on a plane parallel to the substrate, e.g. cross-point arrays
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10B—ELECTRONIC MEMORY DEVICES
    • H10B63/00—Resistance change memory devices, e.g. resistive RAM [ReRAM] devices
    • H10B63/80—Arrangements comprising multiple bistable or multi-stable switching components of the same type on a plane parallel to the substrate, e.g. cross-point arrays
    • H10B63/84—Arrangements comprising multiple bistable or multi-stable switching components of the same type on a plane parallel to the substrate, e.g. cross-point arrays arranged in a direction perpendicular to the substrate, e.g. 3D cell arrays
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10B—ELECTRONIC MEMORY DEVICES
    • H10B63/00—Resistance change memory devices, e.g. resistive RAM [ReRAM] devices
    • H10B63/80—Arrangements comprising multiple bistable or multi-stable switching components of the same type on a plane parallel to the substrate, e.g. cross-point arrays
    • H10B63/84—Arrangements comprising multiple bistable or multi-stable switching components of the same type on a plane parallel to the substrate, e.g. cross-point arrays arranged in a direction perpendicular to the substrate, e.g. 3D cell arrays
    • H10B63/845—Arrangements comprising multiple bistable or multi-stable switching components of the same type on a plane parallel to the substrate, e.g. cross-point arrays arranged in a direction perpendicular to the substrate, e.g. 3D cell arrays the switching components being connected to a common vertical conductor
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10N—ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10N70/00—Solid-state devices having no potential barriers, and specially adapted for rectifying, amplifying, oscillating or switching
    • H10N70/011—Manufacture or treatment of multistable switching devices
    • H10N70/041—Modification of switching materials after formation, e.g. doping
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10N—ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10N70/00—Solid-state devices having no potential barriers, and specially adapted for rectifying, amplifying, oscillating or switching
    • H10N70/011—Manufacture or treatment of multistable switching devices
    • H10N70/041—Modification of switching materials after formation, e.g. doping
    • H10N70/043—Modification of switching materials after formation, e.g. doping by implantation
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10N—ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10N70/00—Solid-state devices having no potential barriers, and specially adapted for rectifying, amplifying, oscillating or switching
    • H10N70/20—Multistable switching devices, e.g. memristors
    • H10N70/24—Multistable switching devices, e.g. memristors based on migration or redistribution of ionic species, e.g. anions, vacancies
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10N—ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10N70/00—Solid-state devices having no potential barriers, and specially adapted for rectifying, amplifying, oscillating or switching
    • H10N70/801—Constructional details of multistable switching devices
    • H10N70/821—Device geometry
    • H10N70/826—Device geometry adapted for essentially vertical current flow, e.g. sandwich or pillar type devices
    • H—ELECTRICITY
    • H10—SEMICONDUCTOR DEVICES; ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10N—ELECTRIC SOLID-STATE DEVICES NOT OTHERWISE PROVIDED FOR
    • H10N70/00—Solid-state devices having no potential barriers, and specially adapted for rectifying, amplifying, oscillating or switching
    • H10N70/801—Constructional details of multistable switching devices
    • H10N70/881—Switching materials
    • H10N70/883—Oxides or nitrides
    • H10N70/8833—Binary metal oxides, e.g. TaOx

Definitions

  • the present invention generally relates mainly to resistive random- access memory devices. More particularly, some embodiments relate to resistive random-access memory formed with pathway structures such as implanted and/or radiated channels for enhancing mobility of charged species, without the need for a forming voltage.
  • a switching layer typically a transition metal oxide is positioned between a top electrode and bottom electrode.
  • the bulk switching layer is initially non-conducting.
  • a sufficiently large voltage a "forming voltage"
  • the forming voltage which generally depends on material quality and thickness of the switching layer(s) can equal the breakdown voltage and can range from a few volts to tens of volts.
  • the conduction paths may be reset (broken, resulting in high resistance) or set (re-formed, resulting in lower resistance) by an appropriately applied voltage (the "switching voltage").
  • the switching voltage the voltage applied to the conduction paths.
  • Some of the disclosed embodiments comprise a resistive random- access memory device that has a first electrode, a second electrode, and a switching region that is between the first and second electrodes and comprises one or more enhanced mobility pathway structures that extend at respective locations in the switching region and are configured to provide enhanced mobility of charged species and have respective electrical resistances that vary with a switching voltage applied between the first and second electrodes, wherein said enhanced mobility pathway structures comprise damage in the switching region caused by ion implantation that includes ions deposited outside the switching region after passing therethrough.
  • the enhanced mobility pathway structures are formed by the ion implantation rather than an application of a voltage across to the switching region equal to or greater than a breakdown voltage.
  • the resistive memory device includes a substrate, wherein the first electrode is a bottom electrode disposed above the substrate and below the switching region and the second electrode is a top electrode disposed above the switching region.
  • the majority of ions used in the ion implantation are deposited below the switching region.
  • a majority of ions used in the ion implantation are deposited below the bottom electrode.
  • At least some of the damage in the switching region is due to collision events caused by the ion implantation, and the majority of the collision events occur below the switching region.
  • the ion implantation causes substantial collision events throughout a thickness of the switching region.
  • the resistive memory device incudes a barrier layer between the switching layer and the first electrode, which barrier layer is functionally disrupted by the ion implantation.
  • the resistive memory device includes a barrier layer between the switching region and the second electrode, which barrier layer is functionally disrupted by the ion implantation.
  • the ion implantation is through the first and second electrodes.
  • the second electrode is above the switching region and comprises a first portion through which said ion implantation takes place and a thickening portion formed subsequently.
  • the resistive memory device includes an underlying substrate having a primary planar upper surface, wherein a first planar interface between the first electrode and the switching region and a second planar interface between the switching region and the second electrode are non-parallel with the primary planar upper surface of the underlying substrate.
  • the first and second planar interfaces are approximately normal to the primary planar upper surface of the underlying substrate.
  • the ions are implanted primarily in a direction that is normal to the primary planar upper surface of the underlying substrate.
  • the resistive memory device comprises defects in the switching region caused by the ion implantation, which defects are approximately uniformly distributed across a plane normal to the first and second planar interfaces.
  • the ion implantation is primarily in a direction that is at an acute angle to the primary planar upper surface of the underlying substrate.
  • the resistive memory device includes at least one of (i) a barrier layer formed between the switching region and the first electrode and functionally disrupted by said ion implantation and (ii) a barrier layer formed between the switching region and the second electrode and functionally disrupted by said ion implantation.
  • the switching region is formed of a transition metal oxide material.
  • the implanted ions are selected from a group consisting of: Ag, Ti, Ta, Hf, O, N, Au, Fe, Ni, Ti, Ta, V, Pb, Bi, W, H, Ar, C, Si, B, P, Ga, As, Te, Al, Zn, In, and Sn.
  • the switching region is formed in situ by said ion implantation.
  • the implantation includes implanting oxygen ions into the switching region, wherein the switching region is initially formed of one or more transition metal materials.
  • the switching region has a built-in stress and the ion implantation increases the stress in the switching region and facilitates creation of said enhanced mobility pathway structures.
  • the switching region comprises at least one of large atomic mass atoms and mismatched material layers causing said built-in stress.
  • the enhanced mobility pathway structures in the switching region are formed at least in part by collision events in said ion implantation, which collision events are distributed substantially uniformly throughout the switching region.
  • Some embodiments include a method of fabricating a resistive random- access memory device comprising forming a first electrode, forming a switching layer, forming a second electrode such that the switching layer is between the first and second electrodes, and carrying out ion implantation in which a portion of the ions both enter and exit the switching layer and thereby facilitate forming one or more enhanced mobility pathways structures that provide enhanced mobility of charged species such that resistance between said first and second electrodes through the switching layer can be increased and decreased by applying a switching voltage between the first and second electrodes.
  • the forming of one or more enhanced mobility pathway structures is due to the ion implantation rather than an application of a voltage to the switching layer equal to or greater than a breakdown voltage for the switching layer.
  • the first electrode is a bottom electrode that is above a substrate and below the switching layer and said second electrode is a top electrode that is above the switching layer.
  • a majority of the ions entering the switching layer in the ion implantation exit the switching layer.
  • the ion implantation causes collision events and the majority of the collision events occurs in locations below the switching layer.
  • the ion implantation functionally disrupts at least one of a barrier layer formed between the switching layer and the first electrode and a barrier layer formed between the switching layer and the second electrode.
  • a first planar interface between the first electrode and the switching layer and a second planar interface between the switching layer and the second electrode are approximately normal with a primary planar upper surface of an underlying substrate.
  • the implanting of ions is primarily in a direction that is normal to the primary planar upper surface of the underlying substrate.
  • the implanting of ions functionally disrupts at least one of a barrier layer formed between the switching layer and the first electrode and a barrier layer formed between the switching layer and the second electrode.
  • a resistive random-access memory device comprising a first electrode, a second electrode, and a switching region that is between the first and second electrodes and comprises one or more enhanced mobility pathway structures that extend at respective locations in the switching region and are configured to provide enhanced mobility of charged species and have respective electrical resistances that vary with a switching voltage applied between the first and second electrodes across the one or more pathway structures, wherein said enhanced mobility pathway structures comprise substantially uniform damage in the switching region caused by electromagnetic irradiation directed into the switching region.
  • electromagnetic irradiation includes radiation of a type selected from a group consisting of: x-rays, gamma rays, UV light, visible light, and IR light.
  • electromagnetic irradiation includes Gamma or X-rays having energy ranging from about 3 KeV to about 100 KeV.
  • electromagnetic irradiation includes X-rays having energy ranging from about 20 KeV to about 30 KeV.
  • electromagnetic irradiation includes X-rays having energy ranging from about 25 KeV to about 27 KeV.
  • the damage is caused in situ by the irradiation after passage thereof through at least one of the first and second electrodes and the switching layer.
  • the device further comprises a plurality of switching layers and electrodes formed in a stack above said second electrode prior to the electromagnetic irradiation.
  • the damage is formed in at least three switching layers electromagnetically irradiated
  • the device further comprises a radiation barrier layer formed below the first electrode, which radiation barrier layer is configured to protect regions below said barrier layer from exposure to the electromagnetic irradiation.
  • the radiation barrier layer is made of a material selected from a group consisting of W, Ta, Bi, Au and Pb.
  • the device includes at least one of (i) a barrier layer formed between the switching region and the first electrode and functionally disrupted by said electromagnetic irradiation and (ii) a barrier layer formed between the switching region and the second electrode and functionally disrupted by said electromagnetic irradiation.
  • the device includes an underlying substrate having a primary planar surface, wherein a first planar interface between the first electrode and switching region and a second planar interface between the switching region and the second electrode are approximately normal to the primary planar upper surface of the underlying substrate.
  • Some embodiments pertain to a method of fabricating a resistive random-access memory device comprising forming a first electrode, forming a switching region, forming a second electrode such that the switching region is positioned between the first and second electrodes, and damaging the switching region by directing energy into the switching region thereby facilitating forming one or more enhanced mobility pathways structures that provide enhanced mobility of charged species such that resistance between said first and second electrodes through the switching layer can be increased and decreased by applying a switching voltage between the first and second electrodes.
  • the damaging causes collision events and the majority of the collision events occurs in locations other than the switching region.
  • the directed energy is in the form of one or more of ions, electrons, X-rays, gamma rays, light including UV and IR light, and ultrasound.
  • the directed energy passes through at least a portion of the first and a portion of the second electrodes.
  • the directed energy functionally disrupts at least one of a barrier layer formed between the switching region and the first electrode and a barrier layer formed between the switching region and the second electrode.
  • Some embodiments pertain to a resistive random-access memory device comprising a first electrode, a second electrode, and a switching region that is between the first and second electrodes and comprises a multiplicity of enhanced mobility pathway structures that extend at respective locations in the switching region and are configured to provide enhanced mobility of charged species and have respective electrical resistances that vary with a switching voltage applied between the first and second electrodes across the pathway structures, wherein said enhanced mobility pathway structures comprise damage in the switching region caused by directed energy beamed into, passing through, and extending beyond the switching region.
  • the enhanced mobility pathway structures comprise damage in the switching region caused by directed energy in the form of one or more of ions, electrons, X-rays, gamma rays, light including UV and IR light, and ultrasound.
  • the enhanced mobility pathway structures comprise damage in the switching region caused by ion implantation into the switching region.
  • the enhanced mobility pathway structures comprise damage in the switching region caused by directed energy in the form of at least one of X-rays, gamma rays, and electrons.
  • the enhanced mobility pathway structures comprise damage in the switching region caused by directed energy in the form of radiation having wavelengths longer than of X-rays.
  • Some embodiments pertain to a resistive random-access memory device comprising a first electrode, a second electrode, and a switching region that is between the first and second electrodes and comprises one or more enhanced mobility pathway structures that extend at respective locations in the switching region and are configured to provide enhanced mobility of charged species and have respective electrical resistances that vary with a switching voltage applied between the first and second electrodes across the one or more pathway structures, wherein the enhanced mobility pathway structures comprise
  • Some embodiments pertain to a resistive random-access memory device comprising, a stack of first electrodes, a stack of second electrodes, and a stack of switching regions each of which is between a respective pair of one of the first electrodes and one of the second electrodes, wherein each of the switching regions comprises one or more enhanced mobility pathway structures configured to provide enhanced mobility of charged species and having respective electrical resistances that vary with a switching voltage applied across the pathway structures, and wherein said enhanced mobility pathway structures comprise damage in the switching regions caused by directed energy concurrently beamed into and through the switching regions in the stack thereof.
  • the damage to the switching regions in the stack caused by the directed energy is caused by directed energy in the form of ion implantation.
  • the damage to the switching regions in the stack caused by the directed energy is caused by directed energy in the form of radiation in the energy range of 1 -100 KeV.
  • the stack of switching region comprises 2-20 switching regions.
  • Some embodiments of the aforementioned device include an array of laterally spaced stacks each comprising first and second electrodes and switching regions, wherein the laterally spaced stacks are interconnected by electrical lines and form an integrated memory structure.
  • FIG. 1 is a cross section diagram showing basic parts of a resistive random access memory RRAM device, according to some embodiments
  • FIG. 2 is a cross section diagram showing paths of implanted ions in a RRAM device, according to some embodiments
  • FIG. 3 is a diagram showing how a focused ion beam can be used to selective area ion implant (SAN) in a RRAM device, according to some
  • FIG. 4 is a diagram showing blank ion implantation in combination with an implant mask, for an RRAM device, according to some embodiments;
  • FIG. 5 is a diagram showing basic parts of a RRAM device under a blanket ion implant without any mask to define a selected area for implantation, according to some embodiments;
  • FIG. 6 is a diagram showing a memory device with multiple ion implantations through an implant mask, according to some embodiments
  • FIG. 7 is a diagram showing a resistive memory device with crystalline and/or almost crystalline and/or polycrystalline switching layer with an implant mask for selective area ion implantation, according to some embodiments;
  • FIG. 8 is a diagram showing a crossbar resistive memory architecture, according to some embodiments.
  • FIG. 9 is a diagram showing stacked RRAM devices having ion implanted paths, according to some embodiments.
  • FIG. 10 is diagram that represents a creation of a mesh network of damage pathways, according to some embodiments.
  • FIG. 1 1 is a diagram showing that the conducting channel formed in a mesh network of damage, according to some embodiments.
  • FIG. 12 is a diagram showing a switching layer that includes multiple layers, according to some embodiments.
  • FIG. 13 is a diagram showing a diode layer that is formed on a switching layer after implantation, according to some embodiments.
  • FIG. 14 is a diagram showing an intentionally damaging implant or radiation occurring on a completed RRAM device, according to some
  • FIGs. 15A and 15B are diagrams showing ion implanting with greater numbers of stacked devices, according to some embodiments.
  • FIG. 16 is a diagram illustrating a switching layer incorporated with large atomic mass atoms that create material stress, according to some embodiments
  • FIG. 17 is a diagram illustrating stress defects caused by mismatched material layers, according to some embodiments.
  • FIGs. 18A and 18B are diagrams illustrating subjecting switching layers that have built-in stress to ion implantation and/or radiation, according to some embodiments;
  • FIG. 19 is a flow chart illustrating basic processing blocks for a basic RRAM, according to some embodiments.
  • FIGs. 20A and 20B are schematic diagrams showing ion implantation doping vs. depth profile for to RRAM devices, according to some embodiments.
  • FIGs. 21 A and 21 B are diagrams illustrating in-situ formation of the switching layer(s) by ion implantation, according to some embodiments.
  • FIGs. 22A and 22B are diagrams illustrating another example of in-situ formation of switching layer(s), according to some embodiments.
  • FIGs. 23A and 23B are diagrams illustrating in-situ formation of switching layers in combination with non-switching layers, according to some embodiments.
  • FIG. 24 is a diagram illustrating ion implanting with a phase change memory device, according to some embodiments.
  • FIGs. 25A and 25B are diagrams illustrating intense ultrasound and/or heat used to generate defects in switching layers, according to some
  • FIG. 26 is a diagram illustrating the use of ion implantation or other radiation to create paths or channels of defects in the switching layer structures, according to some embodiments.
  • FIG. 27 is a diagram illustrating a forming process as is known in the prior art.
  • FIG. 28 is a diagram illustrating variability of switching parameters, according to known RRAM forming processes.
  • FIGs. 29A and 29B are diagrams illustrating aspect of RRAM processing, according to some embodiments.
  • FIGs. 30A and 30B are diagrams illustrating forming a thin TE for ion implantation followed by forming a thicker TE, according to some embodiments;
  • FIGs. 31A and 31 B are diagrams illustrating a RRAM structure with top and bottom electrodes made of a combination of metals and semiconductors, according to some embodiments;
  • FIG. 32 is a flow chart illustrating processing blocks for a basic RRAM, according to some embodiments;
  • FIGs. 33A-33D are plots illustrating simulated result of Hf ions implanted according to a known technique
  • FIG. 34 is a diagram illustrating relative thicknesses of an example of a simple RRAM structure, according to some embodiments.
  • FIGs. 35A-35D are plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 34 at 600 KeV, according to some embodiments;
  • FIGs. 36A-D are plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 34 at 800 KeV, according to some embodiments;
  • FIGs. 37A-D are plots illustrating simulation results of Xe ions implanted into a structure as shown in FIG. 34 at 1600 KeV, or triply ionized and accelerated at 533 KeV, according to some embodiments;
  • FIGs. 38A-D are plots illustrating simulation results of Bi ions implanted into an example structure as shown in FIG. 34 at 3200 KeV, according to some embodiments;
  • FIG. 39 is a log-log plot illustrating the relationship between X-ray linear absorption coefficient in Silicon (in cm -1 ) and X-ray photon energy (in Kev), for energies between 1 Kev and 100 Kev.
  • the use of high dose electromagnetic irradiation, such as X-rays, are shown to facilitate RRAM manufacture, according to some embodiments;
  • FIG. 40 is a diagram illustrating a simple cross section of a RRAM structure that has the switching layer(s) normal or nearly normal to the substrate, according to some embodiments;
  • FIG. 41 is a diagram illustrating a cross section of a RRAM structure where the ion implant or other radiation is performed normal or almost normal to the substrate, according to some embodiments;
  • FIG. 42 is a diagram illustrating a stacked RRAM structure with two or more stacks where the plane of the switching layers are normal and/or almost normal to the surface of the substrate, according to some embodiments;
  • FIG. 43 is a diagram illustrating a cross section of an example of a basic simplified RRAM structure with vertical and/or almost vertical switching layer(s) with ion implantation and/or radiation impinging edgewise into the switching layers, according to some embodiments;
  • FIG. 44 is a diagram illustrating a simple RRAM structure where the switching layer that includes ZnO nanorods, according to some embodiments.
  • FIG. 45 is a diagram showing a structure of multi-layer of RRAM devices according to some embodiments.
  • FIGs. 46A-46H are plots illustrating simulation results of Ar and H ions implanted into a structure as shown in FIG. 45 at various energies, according to some embodiments;
  • FIGs. 47A-47B are expanded plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 45 at 800 KeV, according to some embodiments;
  • FIGs. 48A-48C are expanded plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 45 at 1600 KeV, according to some embodiments;
  • FIGs. 49A-49C are expanded plots illustrating simulation results of H ions implanted into a structure as shown in FIG. 45 at 200 KeV, according to some embodiments;
  • FIG. 50 is a diagram showing a structure of multi-layer of RRAM devices according to some embodiments.
  • FIGs. 51A-51 E are plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 50 at 600 KeV, according to some embodiments;
  • FIGs. 52A-52E are plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 50 at 800 KeV, according to some embodiments;
  • FIG. 53 is a diagram showing a structure of vertically stacked RRAM devices according to some embodiments.
  • FIG. 54 is a diagram showing structure of vertically stacked RRAM devices that is simplified for purposes of simulation, according to some
  • FIGs. 55A-55B are plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 54 at 17.5 MeV, according to some embodiments;
  • FIGs. 56A-56B are plots illustrating simulation results of H ions implanted into a structure as shown in FIG. 54 at 540 KeV, according to some embodiments;
  • FIG. 57 is a diagram showing a structure of vertically stacked RRAM devices that is simplified for purposes of simulation, according to some other embodiments.
  • FIGs. 58A-58D are plots illustrating simulation results of Ar and H ions implanted into a structure as shown in FIG. 57 at 17 MeV and 520 KeV
  • FIG. 59 is a diagram showing structure for RRAM devices used to investigate sensitivity of vacancy concentration in switching layers as a function of top electrode thickness variation due to manufacturing tolerances;
  • FIGs. 60A-60C are plots illustrating simulation results of Ar ions implanted at 17 1600 KeV into a structure as shown in FIG. 59 having top electrode 5940 thicknesses of 80, 100 and 120 nm, according to some
  • FIGs. 61 A and 61 B illustrate constrictions in vacancy filaments
  • FIG. 62A is a diagram illustrating a uniform distribution of the vacancies within the switching layer, according to many of the embodiments described herein;
  • FIG. 62B is a diagram illustrating vacancy distribution resulting from a known technique of ion implanting prior to the formation of the top electrode
  • FIGs. 63A-63D are diagrams showing a simplified sequence of processes for manufacturing a portion of RRAM devices, according to some embodiments.
  • FIG. 64A-64B are diagrams illustrating ions knocked into an SL layer by ion implantation through a TE layer, according to some embodiments.
  • FIGs. 65A-65B are diagrams illustrating ions knocked into an SL layer with a diode, semiconductor, and/or tunnel layer deposited, according to some embodiments;
  • FIGs. 66A-66B are diagrams illustrating ions knocked into an SL layer with a diode, semiconductor, and/or tunnel layer, and TE deposited, according to some embodiments;
  • FIG. 67 is a simplified flow chart of the process of implanting through a TE, according to some embodiments.
  • enhanced ion/vacancy mobility channels are provided for the resistive random access memory, memristor, and phase change memory (electron and hole paths) for a more definitive conduction path rather than more random conductive paths resulting from applied forming voltages.
  • a path or paths are provided for the formation of filaments and/or ionic movements and hopping of the defects and ions.
  • ions are accelerated and implanted into the switching layer and through the switching layer.
  • US2008/0090337A1 top and bottom electrode sandwiching a multiple layered switching layer and a crossbar array memory architecture
  • US2004/0160812A1 two terminal resistive memory device
  • US2008/0185572A1 memristive or resisitve random access memory (RRAM) two terminal device and a crossbar architecture
  • US2008/0296550A1 multilevel stacked crossbar architecture using the two terminal RRAM.
  • Memristive memory and RRAM are similar.
  • the initiating of the filaments relies predominantly on a forming process where a relatively high voltage (significantly higher than the operating voltage) is used to cause a "breakdown" and/or create a defect path in the switching layer.
  • This forming voltage approaches the breakdown voltage and is dependent on switching layer thickness and quality.
  • the breakdown field of oxides ranges between 5x10 6 V/cm and 8x10 6 V/cm.
  • HRS high resistive state
  • LRS low resistive state
  • memristor/RRAM multiple HRS and LRS states are possible where the IV (current voltage) characteristics are highly non-linear and/or there are multiple bi-stable states.
  • IV current voltage
  • HRS/LRS are observed without a forming process, there may exist a defect path where ions, vacancies (also called species) can traverse to form filaments for switching characteristics, for example; however these are observed rarely and not reproducible which is not conducive to high volume manufacturing.
  • ions, vacancies also called species
  • some memory devices may require a forming process and some may not.
  • a path is introduced for the formation of filaments and/or ion/vacancies/pint defect diffusion/drift/hopping such that the forming process using high voltage need not be relied upon.
  • the path or channel is formed by ion implantation where the passage of the accelerated ions such as Au, Pt, Pb, Fe, Bi, U, Ar, O, N, Si, B, P, H, Ti, Ta, C, Zr, Zn, Ga, In, Se, through the switching layer(s) causes defects, including crystallographic defects such as vacancies, interstitial defects, point defects, and lattice defects to form in the crystalline or noncrystalline switching layer.
  • the ions can be implanted into the switching layer(s) to form a conducting channel of ions such as Au, Pt, C, Bi, Zr, In, Ga to name a few, and / or the damage to the crystal or micro-crystal (in the case of amorphous switching layer(s)) forming crystallographic defects that can also assist in the formation of filaments resulting in a HRS and LRS memory device.
  • Advantages of ion implantation include that it is a manufacturable process, and that it generates nearly one-dimensional defect paths that are approximately the shortest distance between the top and bottom electrodes.
  • Filaments can be single wire-like or tree-like with multiple paths feeding to a main trunk or trunks. Single or multiple filaments can thus be approximately one-dimensional and/or three-dimensional.
  • Known RRAMs use forming processes that rely on applying a forming voltage, which is comparable to the breakdown voltage and/or approximately ten times greater than the operating voltage, to the switching layer and/or impurities at the interfaces of the switching layer(s) and the electrodes.
  • the use of such forming voltages is necessary for known RRAM devices to exhibit bi-stable and/or multi- stable HRS and LRS states.
  • the application of high forming voltages can be avoided. Instead, an initiation voltage can be applied that may be higher than the operating voltage in order to initiate the bi-stable and/or multi-stable HRS and LRS states.
  • the initiation voltage may be a few tenth of a volt to a few volts higher than the operating voltage, which is less than the breakdown voltage.
  • Silicon dioxide has a breakdown field of 6x10 6 V/cm and hafnium oxide has a breakdown field of approximately 5x10 6 V/cm.
  • FIG. 1 is a cross section diagram showing basic parts of a resistive random access memory RRAM device, according to some embodiments.
  • RRAM device 100 includes top electrode (TE) 140, switching layer(s) 130 and bottom electrode (BE) 120.
  • Paths of crystallographic defects can be created using ion implantation beam of ions 150 with energies that penetrate the entire structure of RRAM 100 and at least some come to a rest in the substrate 1 10.
  • Defects created by the ion beam 150 include vacancies, interstitials, point defects, line defects, plane defects, Frenkel defects, antisite defects, impurity, amorphous, topological defects, dislocations, F-centers, to name a few.
  • These crystallographic defects create a channeling path that can be conducive for ions/electrons and/or vacancies to migrate more easily under an electric field bias.
  • Ions used in beam 150 can include one or more of Au, Pb, U, Er, Yt, Zn, O, Pt, Fe, Bi, H, He, Ga, Hf, N, C, V, Ti, Ta.
  • FIG. 2 is a cross section diagram showing paths of implanted ions in a RRAM device, according to some embodiments.
  • the path 250 of the implanted ions creates crystallographic defects in crystalline and/or non-crystalline switching layers 130. Such defects can be vacancies, point defects, and/or interstitial defects.
  • the accelerated ions are implanted at or to a certain depth with a certain spatial distribution.
  • the ions can be implanted into one or more of the switching layer 130, electrodes 140 and 120, and the substrate 1 10.
  • the implanted ions shown in FIG. 2 are implanted such that they come to rest in region 252 of substrate 1 10.
  • Accelerated ions in ion implantation will scatter and there will be a distribution of scattered ions spatially. However, based on the acceleration voltage, the ions will propagate generally in their ballistic path with the maximum scattering occurring when the accelerated ions are coming to rest in the material. Dose is another factor that can determine scatter. Dose can range from 1 .0x10 6 /cm 2 to 1 .0x10 16 /cm 2 or higher depending on the implant depth which is dependent on the acceleration energy, ion mass, and the mass of the material that is being implanted into.
  • Acceleration energies can range from less than 1 Kev (kilo electron volt) to 3200 KeV or more depending on the depth of implantation and mass of the ions and target density.
  • the Au ions can be ion implanted with an accelerating energy of approximately 4.8 kilo-electron volt (KeV) per nm (nanometer) of depth in a crystalline material such as AI2O3. See,
  • Dose can be adjusted to achieve the optimal concentration of defects for desired IV characteristics for a particular application. For example, doses as high as 5x10 16 ions/cm 2 or higher to create a region of very highly concentrated defects, resulting in extremely low switching voltages. With less defect concentration (at a lower implantation dose of Au ions for example), the switching voltage may be higher.
  • FIG. 3 is a diagram showing how a focused ion beam can be used to selective area ion implant (SAN) in a RRAM device, according to some
  • a focused ion beam 350 is used to form a narrow defect path/channel 354 in the switching layer 130.
  • switching layer 130 can be crystalline, or it can be non-crystalline such as amorphous while still having microcrystalline structures for example.
  • Defect path/channel 354 facilitates the movement of the ions, vacancies, and/or point defects, for example, so as to form filaments that can result in RRAM uniform switching characteristics without the use of a high forming voltage.
  • the ion beam can penetrate the TE 140 into the switching layer 130 and into the BE 120 and into the substrate 1 10.
  • FIG. 4 is a diagram showing blank ion implantation in combination with an implant mask, for an RRAM device, according to some embodiments.
  • Blanket ion implantation is generally more manufacturable than using a focused ion beam.
  • the implant mask 410 can be Au, polymer, SiOx for example. In general, any material whose thickness is enough to contain most of the implanted ions is suitable. Typically, a dense material, high atomic number, such as Au will require a thinner layer than a less dense material, low atomic number, to stop most of the implanted ions.
  • An opening 412 in the mask 410 allow the ions to pass through and implant into the desired regions of the RRAM device 100 depending on the accelerating energy.
  • the amount of ions implanted is the dose.
  • ions are implanted into the substrate 1 10, with their paths going through the top electrode 140 and the switching layer 130 and generating defects 454.
  • the defects assist the diffusion/hopping/drift of ions and vacancies in the formation of filaments such that switching can occur uniformly without the use of a forming voltage. This pre-existence of the defect paths/channels reduces the randomness of the formation of the filament that would otherwise require a forming or breakdown voltage.
  • FIG. 5 is a diagram showing basic parts of a RRAM device under a blanket ion implant without any mask to define a selected area for implantation, according to some embodiments. Approximately parallel paths/channels are created which lead to more uniform formation of filaments without relying on a forming voltage to initially form filaments and/or establish switching current -voltage (IV) characteristics. As shown in FIG. 5, the ions are implanted into the region 552 of substrate 1 10 for example. Implanting the ions partially or completely within any of the layers comprising the RRAM device (TE, Switching layer, BE, substrate) are also possible, for example. Note that ion beam 550 can be normal to the surface of top electrode 140 or it can be at an angle, according to some embodiments.
  • FIG. 6 is a diagram showing a memory device with multiple ion implantations through an implant mask, according to some embodiments.
  • a blanket implant without an implant mask 410 can also be used. Ions are implanted into region 652 of the substrate 1 10 for creation of a defect path in the crystalline, non-crystalline, or amorphous
  • the implant process used to create the defect paths in the switching layer can include multiple implants of different ions. Examples of different ions include Au, Pt, Pb, H, O, N, C, Ar, Ti, Ta, Ga, and In.
  • ion implants of O, N, Ti, Ta can be implanted partially or entirely within the switching layer 130 for example in region 654, to provide ions for switching/oxidation and reduction in a TiOx switching layer resistive memory device.
  • the order of the implantation for generating defects and for implanting ions to assist in the switching process may be interchanged. Ion implantation can be through the top electrode 140 and into the switching layer 130 and into the bottom electrode 120 and to the substrate 1 10.
  • FIG. 7 is a diagram showing a resistive memory device with crystalline and/or almost crystalline and/or polycrystalline switching layer with an implant mask for selective area ion implantation, according to some embodiments. Ions of sufficient energy are implanted through the structure and into the substrate 1 10 and/or bottom electrode 120 with sufficient dose to cause defect paths that are mostly normal to the surface. The resulting defect paths can be considered as amorphous due to the damage caused by the implanted ions traversing a crystalline and or almost crystalline switching layer(s) 730.
  • ions/vacancies/point defects/defects/oxidation-reduction, and electrochemical processes can more easily redox/drift/diffuse/hop under an applied electric field in the damaged region 754 rather than in the undamaged regions of layer 730 that are crystalline/poly-crystalline and/or almost crystalline or regions where it is stoichiometric and the damaged region together with or without added implants of ions that participate in the filament process, such as O, H, Ti in a TiOx switching layer which are less stoichiometric.
  • the crystalline switching layer 730 can be T1O2 stoichiometric and the implant damaged region 754, with or without O or Ti ions, are TiOx are non-stoichiometric.
  • the non-stoichiometric region 754 can facilitate formation of filaments and bi-stable/ multi-stable states of HRS and LRS.
  • the stoichiometric regions may be less susceptible to dielectric breakdown at the high field regions occurring at the edge of the electrodes which improve reliability of the RRAM. Switching activity will take place away from the edge and mostly at the regions where the ion implant ions caused defect and crystal damages, such as break in lattice, vacancies, point defects, interstitial, to name a few.
  • FIG. 8 is a diagram showing a crossbar resistive memory architecture, according to some embodiments. Each intersection point of the horizontal and vertical electrodes represent a resistive memory device.
  • the crossbar memory architecture 800 is fabricated and the ion implantation and or radiation can be a blanket implant without an implant mask and/or it can have an implant mask for selective area ion implantation and or radiation.
  • the switching layer can be implanted where only the BE and TE are formed. If the BE is Ti, O ions can be implanted partially into the BE to form a TiOx switching layer, for example.
  • TiN can be the BE, and Hf and O can be implanted partially into TiN to form HfOx where the top electrode can be Mo, Pt, Au, Ti, to name a few.
  • the BE can be TiN and Hf oxide grown by atomic layer deposition (ALD), for example, followed by a top electrode of Mo.
  • ALD atomic layer deposition
  • Selective area ion implantation at regions away from the edge of the device, where high fields are concentrated due to electrode edges, can increase the reliability of the resistive memory devices. Using selective area ion implantation with an implant mask, regions away from the device edges can be implanted with paths/channels for filament formation and with implant of O, H, Hf, Ti ions for example.
  • the implantation assists in filament formation via defects such as vacancies hopping and/or drift under an applied electric field, and assists oxidation-reduction processes.
  • Vacancies can also be formed in non-transition metal oxide materials such as N, C, F in SiNx, SiCx, florides, that are based on nitride, TiN, AIN, SiN, for example, carbide, SiC, WC, NbC, TiC for example, fluorides, CaFx, MgFx, WFx for example other than oxides or in any combinations of oxides such as nitrides, fluorides, carbides, to name a few.
  • non-transition metal oxide materials such as N, C, F in SiNx, SiCx, florides, that are based on nitride, TiN, AIN, SiN, for example, carbide, SiC, WC, NbC, TiC for example, fluorides, CaFx, MgFx, WFx for example other than oxides or
  • FIG. 9 is a diagram showing stacked RRAM devices having ion implanted paths, according to some embodiments.
  • the stacked device
  • the architecture 900 can, for example, be implemented using crossbar resistive memory devices.
  • Selective area ion implantation or non-selective area ion implantation can be used simply by increasing the acceleration energy of the ions to be implanted to penetrate to the desired depth, for example as shown, the defect path channel 954 is generated by ions that are implanted to the substrate 1 10, penetrating both switching layers 930 and 130. Note that more than two switching layers can also be penetrated by the implanted ions by adjusting the acceleration energy.
  • Typical commercial ion implant equipment can accelerate at 600 kilovolts (KV) and by using higher ionization number such as doubly ionized or triply ionized ions, the acceleration energy of the ions can be doubled or tripled.
  • ions can be implanted entirely and/or partially in RRAM#1 (made of TE 940, SL 930 and BE 920) and RRAM#2 (made of TE 140, SL 130 and BE 120). Radiation can also be used in place of or together with ion implantation,
  • ions are distributed in depth and width in a targeted material.
  • the peak distribution of the implanted ions can be in the substrate for example and still have some straggling ions in the switching layers and the TE and BE, for example.
  • crystalline also can include “stoichiometric”, such that wherever a crystalline region is mentioned it can refer to a stoichiometric region.
  • the damage may or may not be orderly. This is also true for any other methods of damaging the crystalline, non-crystalline, stoichiometric and/or non-stoichiometric material such as proton implantation, electron beam, any particle beam, x-ray, gamma ray, optical beam, to name a few.
  • any other methods of damaging the crystalline, non-crystalline, stoichiometric and/or non-stoichiometric material such as proton implantation, electron beam, any particle beam, x-ray, gamma ray, optical beam, to name a few.
  • the damage paths are shown schematically as straight lines for simplicity. In practice, the damage can be a network of
  • FIG. 10 is diagram that represents a creation of a mesh network of damage pathways, according to some embodiments. Collisions of implanted particles with the material into which the particle is implanted can create a dense mesh network of intersecting paths as shown.
  • the particles or energy can be ions, protons, electrons, X-ray, gamma ray, light from deep ultraviolet (UV) to infrared (IR), to name a few.
  • the switching layer can consist of multiple layers for switching and for nonlinear electrical conduction such as a diode and other resistive and semiconducting layers for optimum memory and/or logic operations.
  • a shortest path within the switching layers can be found to form conduction channels for the filaments or channels for filament formation that eliminate the necessity of forming process also known as breakdown process.
  • Such forming/breakdown processes often exceed 5-20 volts and can reach 30 volts or more depending on the thickness of the switching layer.
  • Breakdown field of oxides are typically 5x10 6 V/cm to 6x10 6 V/cm. Breakdown voltage also depends on the quality of the oxide. Oxides with higher concentrations of defects and/or grains have effectively less thickness of high-quality oxide and therefore have lower breakdown voltages. Since shortest path can always be found, electrically and/or electrochemically, the IV characteristics have less variations.
  • FIG. 1 1 is a diagram showing a conducting channel formed in a mesh network of damage, according to some embodiments.
  • a conducting channel 1 154 can be formed along a short path along the damage area that may not be a perfectly straight line. However, this is intentional damage and not a breakdown process, such as the forming process in known RRAM designs. Since the intentionally damaged volume is a dense mesh network, the shortest path through the switching layer(s) can be approximated by a straight line, as shown by the pathway 1 154 that basically connects the shortest or nearly shortest distance of pathways shown in dotted lines.
  • FIG. 12 is a diagram showing a switching layer that includes multiple layers, according to some embodiments.
  • Switching layer 130 can include multiple layers that can be stoichiometric and/or non-stoichiometric, crystalline, noncrystalline and/or micro-crystalline.
  • Switching layer 130 can also include layers that are not involved in switching.
  • non-switching layers can include nitride layers, semiconducting layers such as amorphous silicon, carbide layers, chloride and fluoride layers to name a few. These layers can be implanted first to create intentional damage such that conduction channels can be formed along a shortest path without breakdown voltages.
  • Species such as oxygen ions, nitrogen, H, Ar, C, CI, F, Ne, He, Xe ions , electrons, metal ions such as Au, Pt, Ag, Fe, Ni, In, Ga, Zr, Ta, Ti, Zn, Sn, Sb, Hf, Er, Yb, Gd, Pb, Bi, W, Zr, Os, can be implanted into the switching layer(s) and/or into the electrodes and substrates. Ions can be either for doping purposes, for example to create an oxygen rich reservoir and or vacancies for electro-chemical reactions in the case of transition metal oxide switching layer(s), and ions can be used to damage the material intentionally to create paths for the formation of filaments.
  • Ions can also be used to change materials, for example, in the case of transition metal electrodes such as Ti, Ta, Mo, W, V, Zr, Ni, Fe, oxygen ions can be implanted partially and/or completely into the electrode to convert the transition metal electrode to transition metal oxide for switch layer(s).
  • This in situ formation of the switching layer(s) avoids contamination as may occur in the case when the switching layer is formed separately and is exposed to atmosphere before the top or subsequent electrodes are formed to complete the RRAM device. Thermal treatment may or may not be desirable or necessary to optimize the in situ switching layer.
  • FIG. 13 is a diagram showing a diode layer that is formed on a switching layer after implantation, according to some embodiments.
  • the switching layer 130 is implanted with ions, particles, protons, electrons, rays such as x-rays, gamma rays, light waves such as ultraviolet, blue, green, red, infrared wavelengths to cause damage by stress, heat, uneven thermal expansion, or radiation.
  • damage can be caused by insertion of large atomic mass atoms in the material by co-sputtering or co-deposition.
  • a diode layer or layers 1310 and any non-switching layers to control current flow such as resistive layers can be grown on top of the intentionally damaged layer to have good diode and/or resistive characteristics.
  • a top electrode completes this simplified RRAM. Further ion implantation and or radiation processing can be used to alleviate any unintentional barrier layers between the top electrode 140 and the diode layer 1310 and/or the diode layer 1310 and the switching layer 130.
  • FIG. 14 is a diagram showing an intentionally damaging implant or radiation occurring on a completed RRAM device, according to some
  • diode layer 1310 can be amorphous silicon and does not participate in the transition metal oxide switching process where the switching layer(s) can be composed of transition metal oxide(s). This greatly simplifies manufacturing since the crossbar can be first made and the intentional damage introduced after by exposure to ions, particles, protons and/or radiation singularly or in any combinations of species, energies, dose and methods.
  • FIGs. 15A and 15B are diagrams showing ion implanting with greater numbers of stacked devices, according to some embodiments.
  • the crossbar architecture such as shown in FIG. 15A can be stacked from two to over twenty crossbars on top of each.
  • FIG. 15B shows a stack of devices with a single ion implant path 1554.
  • the average thickness of each crossbar is between about 0.1 to 2 microns.
  • Stacking twenty or more crossbars will result in a thickness ranging approximately from 2 to 40 microns.
  • Ions and electrons can be accelerated from about KeV (kilo electron volt) to MeV (mega electron volt) to penetrate the stack of RRAM devices.
  • Approximately 1 -10 MeV protons can penetrate more than tens of microns of material depending on the material.
  • penetration can be approximately 25 to 250 microns using 1 MeV and 10 MeV respectively.
  • penetration is
  • ion implantation is performed layer by layer as the layers are deposited and patterned. In this case only about 0.1 -1200 KeV acceleration energies are necessary for ion/particle implantation and/or exposure to radiations such as x-ray, gamma ray, UV to IR.
  • FIG. 16 is a diagram illustrating a switching layer incorporated with large atomic mass atoms that create material stress, according to some embodiments.
  • Switching layer(s) 1630 can be engineered by including atoms, such as atom 1610, with large atomic mass within the switching layer(s) 1630 that causes local strain in the material resulting in defects.
  • the large atoms can be incorporated by any of several methods. In one method, while depositing the switching layer(s), large atomic mass atoms such as Au, Pb, Bi, Ga, are co-deposited simultaneously with depositing the other material of layer 1630.
  • the large atomic mass atoms can be in atomic form or in cluster form such as when deposited by sputtering.
  • An example is co-sputtering both TiOx and Pt or Au, such that Au or Pt atoms and/or clusters are embedded in the TiOx switching layer(s).
  • This can cause defects within the switching layer(s) resulting in a network of paths for conducting channels for oxygen ions, vacancies, and electrons to propagate.
  • the channels form filaments, including through oxidation and reduction, where vacancies and ions can cause transition metal oxides to be more or less conducting, depending on the amount of oxygen ions and vacancies.
  • TiOx with high oxygen ions and low vacancies are less conducting than TiOx with less oxygen ions and more vacancies.
  • Other switching layer(s) such as HfOx, VOx, TaOx, WOx and GaOx or PbOx can be co-deposited or sequentially deposited by ALD and/or PECVD.
  • Ion implantation can also be used to introduce ions into the switching material thereby causing stress and strain resulting in a network of defects.
  • the resulting channels through which charged ions and vacancies propagate form filaments whose resistivity can be changed by the concentration of ions and vacancies.
  • Implanted ions can include O, H, He, Ar, Ne, Xe, Au, Pt, Mn, W, Mo, Zn, Pb, Bi, Ti, Ta.
  • the implanted ions to reside partially in the switching layer(s) they can also reside predominantly outside the switching layer(s) where the ions by transiting the switching layer(s) can cause defects to the materials in the switching layers and other layers that the ions transited.
  • Material layers 1730 can be formed by growing epitaxially or non-epitaxially such that they are mismatched in lattice, thermal expansion, or deposition parameters such as temperature, pressure, stoichiometric, or material composition.
  • layers 1730 can be a superlattice with material layer thicknesses ranging from 0.1 to 3 nm where material A and material B are formed using different deposition methods. Different methods include different deposition parameters, such as temperature, rate, pressure, and/or material composition that result in a stress between material A and material B. In some cases there can be more than two materials interlaced in a repeating or non-repeating pattern.
  • the switching layer(s) can be material AB, ABABABAB or ABCDBDAC, or ABCDEFG to name a few, where each letter represents a material layer different from other material letter in properties due to different deposition parameters, deposition methods, and/or composition of material to name a few .
  • the resulting stress can result in defect networks that can form conducting channels 1750 for ions, vacancies, electrons to form the filaments necessary for memory effects in RRAM.
  • FIGs. 18A and 18B are diagrams illustrating subjecting switching layers that have built-in stress to ion implantation and/or radiation, according to some embodiments.
  • switching layer 1630 has large atoms embedded and in FIG. 18B, switching layer 1730 has a superlattice or layered structures with layers that are different in properties that result in stress such as due to different material composition and/or deposition methods and/or other parameters that result in stress.
  • the structures can be subjected to ion implantation, radiation by UV to IR to further exacerbate the stress and assist in creating defects in the material so that a network of defective paths can form channels for filament formation in the operation of a RRAM without a need to use a breakdown voltage.
  • FIG. 19 is a flow chart illustrating basic processing blocks for a basic RRAM, according to some embodiments.
  • the process starts with a substrate.
  • the substrate includes a layer or layers such as silicon dioxide to insulate the RRAM from the substrate electrically.
  • the substrate can also contain CMOS circuits that can be connected to the RRAM electrically.
  • the bottom electrode is defined and deposited. This can involve metal deposition, such as evaporation of metals or sputtering of metals, photolithography, wet and/or dry etching such as reactive ion etch, ion milling.
  • nanoimprinting is used to define the electrodes.
  • Metals such as Ti, Mo, W, Ni, Pt,AI, Zr, Ta, V and/or metal nitride such as TiN, TaN, GaN, AIN, InN, WN, and silicide such as NiSi, PtSi, WSi can be used.
  • the switching layer(s) is deposited along with any other layers, such as diode layers and/or resistive layers, for current control. As mentioned earlier, ion implant and/or radiation can also be performed in this process block.
  • the switching layer(s), such as TiOx, TaOx, VOx, HfOx, ZnOx, GaOx, InOx, PbOx, FeOx, CrOx, NiOx, to name a few can be stoichiometric and/or non-stoichimetric.
  • the switching layer(s) can be deposited in several methods such as atomic layer deposition (ALD), plasma enhanced chemical vapor deposition (PECVD), sputtering, evaporation, molecular beam epitaxy (MBE).
  • Block 1916 is similar to block 1912. In this case the top electrode is deposited and defined photolithographically and/or by nanoimprinting.
  • the electrodes can be evaporated, sputtered, PECVD, ALD, to name a few.
  • Metals such as Ti, Mo, W, Ni, Pt,AI, Zr, Ta, V and/or metal nitride such as TiN, TaN, GaN, AIN, InN, WN, and or silicide NiSi, PtSi, WSi can be used.
  • ion implantation and/or radiation such as Xray, gamma ray, optical rays from UV to IR are used to intentionally generate defects in the switching layer(s).
  • dopant ions such as oxygen can be implanted to provide excess oxygen ions to affect the resistivity of the filaments.
  • H ions can also be implanted to generate vacancies that can affect the resistivity of the filament.
  • defects are generated using heavy ions that are implanted through the switching layers and/or ions to generate defects.
  • the heavy ions pass through the switching layer(s) and many or most are embedded outside the switching layer(s).
  • Particles implanted can include electrons, ions such as H (proton), O, Ti, Ta, N, F, CI, C, Li, Na, K, He, Ne, Xe, Ar, Au, Be, Mg, Mn, Si, B, Ge, Ga, Ni, V, Au, Pt, Fe, Pb, Bi, In, W, Mo, Hf, Ag, Cr.
  • Implant energies can be from less than about 1 KeV to greater than 10s MeV, and in certain cases to GeV. Doses can range from 1 x10 4 /cm 2 to 1 x10 7 /cm 2 or more. According to some embodiments, multiple ions species, energies and doses are used. According to some embodiments, block 1918 includes radiation by X-rays, gamma rays and/or optical rays from UV to IR. According to some embodiments, an electron beam is used to modify and create defects in materials.
  • the ion implant/radiation process block 1918 can be performed last and/or it can be performed after a certain number of stacks, for example at every 5 stacks. In some cases, block 1918 is performed at every RRAM layer and/or sublayer (where the switching layer(s) are deposited). Note that block 1918 can also include electrons with energies in the sub KeV to hundreds of KeV and MeV and variable doses can range from 1 x10 4 /cm 2 to 1 x10 7cm 2 or more.
  • FIGs. 20A and 20B are schematic diagrams showing ion implantation doping vs. depth profile for two RRAM devices, according to some embodiments.
  • curve 2010 is the ion implant profile for O ions in the switching layer
  • curve 2012 is the ion implant profile for B or P ions embedded mostly in the bottom electrode of a Ti/TiOx/Pt structure.
  • curve 2020 is the ion implant profile for O ions in the switching layer
  • curve 2022 is the ion implant profile for Ar or In ions embedded mostly in the bottom electrode of a TiN/HfOx/Ti structure.
  • FIGs. 21 A and 21 B are diagrams illustrating in-situ formation of the switching layer(s) by ion implantation, according to some embodiments.
  • FIG. 21A shows just the BE 2120 and TE 2140.
  • One or both of 2120 and 2140 is a transition metal.
  • an ion implantation process 2150 implants oxygen ions approximately at the interface and/or predominately into one of the electrodes 2120 and 2140 to form an in-situ transition metal oxide layer 2130, which is the switching layer.
  • multiple switching layers can be formed by using layers of different metal in the electrodes.
  • BE 2120 is made of layers of Pt/Ti/V
  • the TE 2140 is Mo.
  • the resulting switching layers 2130 are TiOx/VOx/MoOx. Note that in this case only part of the TE 2140 is converted to MoOx.
  • ion implantation with other ions can be used, such as Au, Pt, Xe, H, N, CI, F, Si to name a few to either be partially embedded into the switching layers and/or to mostly embedded outside the switching layer to create intentional defects for filament formation.
  • Other in situ switching layers such as TiN, AIN, InN, GaN can be fabricated similarly using N ion implantation for example into electrodes of Ti, Al, or In.
  • FIGs. 22A and 22B are diagrams illustrating another example of in-situ formation of switching layer(s), according to some embodiments.
  • the BE 2220 and TE 2240 need not be transition metals since one or more transition metal layers 2200 and 2202 are sandwiched between them.
  • the bottom and top electrodes 2220 and 2240 are materials such as Pt, Au, Al, Cu, Ag, TiN, GaN, InN, silicides, nitrides, carbides, carbon, and/or semiconductors.
  • the intermediate layers 2200 and 2202 can be transition metals such as Ti, Ta, V, Ni, Hf, Zr, Cr, Zn, Mn, W, Mo, Ir, Os, Al.
  • Oxygen ion implant doses range from approximately 1 x10 0 /cm 2 to 5x10 20 /cm 2 or greater at energies from a few KeV to 800 KeV or greater.
  • Nitride based switching layers can also be fabricated for example layer 2202 can be Ti, In, or Al and N ions are implanted.
  • multiple different kinds of switching layers such as transition metal oxide and transition metal nitride can be fabricated in situ with O and N ion implantation.
  • filament formation refers to the formation of a channel or channels through which electric current flows along material defects, wherein the resistivity of the channel(s) can be modified by charged ions and vacancies. Note that the width of each channel can range from a small fraction of the size of the lateral dimension of the RRAM to the actual lateral dimension of the RRAM.
  • FIGs. 23A and 23B are diagrams illustrating in-situ formation of switching layers in combination with non-switching layers, according to some embodiments.
  • other layers can also be formed.
  • a tunnel silicon dioxide layer 2334 of a few nm thick can be formed by the addition of a Si layer 2304 which can then be oxidized to silicon dioxide and/or SiOx by the oxygen ion implantation 2350.
  • other layers can also be used that instead of forming oxide can change the electrical characteristics of the layer such as GaAs, GaN, InN, to name a few.
  • implanted oxygen ions can increase the resistivity of the layer for current control, and also may have nonlinear IV characteristics than can minimize sneak current in a crossbar RRAM architecture.
  • transition metal oxide based transition metal oxide based
  • the techniques described are applied to other resistive memory devices not based on transition metal oxide for switching, such as transition metal nitride, transition metal carbide, metal sulfides, metal chlorides, metal flourides, sulfides, chlorides, fluorides, diamond-like carbon, carbon, polymer, and phase change non-volatile memory devices.
  • resistive random access memory also known as memristors can also be used in logic operations.
  • the techniques described herein are not limited to memory devices but apply also to logic operations. See, e.g. "NATURE NANOTECHNOLOGY
  • phase change memory Ge is sandwiched between top and bottom electrodes, an ion implant mask is used, as in FIG. 4 for example, for selective area ion implantation.
  • Sb and Te can be implanted such that the phase change material Ge-Sb-Te can be formed in-situ.
  • Oxygen ions can also be implanted in areas where phase change material is not desirable; for example, an ion implant mask can be used in selective area ion implant regions to reduce the size of the phase change area to improve electrical performance, such as low current operation.
  • Other PCM material such as CuS, AgS, Sb-Te, Sb-Te-TiN, to name a few, may also be used, where Cu electrodes can be partially converted to CuS with S ion implantation.
  • FIG. 24 is a diagram illustrating ion implanting with a phase change memory device, according to some embodiments.
  • a Ge-Sb-Te phase change memory region 2452 is formed within a Ge layer 2430 that is sandwiched between a bottom and top electrodes 140 and 120.
  • O ions can first be implanted using an implant mask (not shown) everywhere in the Ge region except where Sb and Te will be implanted.
  • a second implant mask 2410 can be used.
  • the mask 2410 is self-aligned to the first implant mask where the first implant mask is removed exposing an aperture.
  • Implantation process 2450 then implants Sb and Te ions selectively into the Ge forming the Ge-Sb-Te phase change material 2452 for memory completely in-situ. See e.g., MRS BULLETIN ⁇ VOLUME 39 ⁇
  • Heat treatment may or may not be desirable or necessary to optimize the device performance.
  • defect paths or channels are created by ion beam, electron beam, electromagnetic waves, ultrasound, internal stress, and/or built-in stress in the switching material, which can be in crystalline, noncrystalline, microcrstalline and/or amorphous material.
  • the switching material can be in crystalline, noncrystalline, microcrstalline and/or amorphous material.
  • chemical bonds between metal and oxygen are broken microscopically such that vacancies are formed.
  • the vacancies can be, for example, oxygen vacancies and/or metal vacancies.
  • These defects can form a path or channel such that vacancies and ions can drift under the influence of an electric field and/or potential.
  • a path of defects where bonds are broken is created using ion implantation, electron beam, electromagnetic waves, heat, ultrasound, built-in stress.
  • Defects can take the form of vacancies, interstitials, point defects, line defects, plane defects, Frenkel defects, antisite defects, impurities, amorphous regions, topological defects, dislocations, and/or F-centers.
  • Amorphous and polycrystalline materials have microscopic crystal structures and crystalline materials have macroscopic crystal structures.
  • Defects such as bond breakage can provide a path/channel for vacancies, ions, electron diffusion and drift under an electric field and/or potential applied across the top and bottom electrodes of a resistive random access memory (RRAM).
  • RRAM resistive random access memory
  • Filaments can be non-uniform in diameter as a function of length and can have multiple branches. Furthermore, multiple filaments may be formed during a forming process. In polycrystalline switching layers, the defects can form along grain boundaries. By purposely introducing the defective channels and/or paths described herein, the filaments are formed mostly along the purposely introduced defective paths/channels. By using purposely formed defective path/channels, the known forming process under high electric field will be unnecessary. In some cases, an applied voltage to initiate the drift of vacancies, ions, electrons to form the filament may be used instead. This initial voltage, however, will be in the neighborhood of a few volts and less than the bond breaking voltage necessary for forming processes.
  • FIGs. 25A and 25B are diagrams illustrating intense ultrasound and/or heat used to generate defects in switching layers, according to some
  • FIG. 25A shows a switching layer 1630 having large atomic mass atoms incorporated therein which cause local strain in the material.
  • Ultrasound 2550 provides external stimuli to cause local defects to propagate.
  • layers 1730 include mismatched material layers that result in stress between neighboring layers.
  • Ultrasound 2552 provides external stimuli to cause local defects 1750 to propagate.
  • heat is used with or without the ultrasound to provide the external stimuli to propagate the defects.
  • the switching layers do not have built-in internal stress but the ultrasound is intense enough to nevertheless cause and propagate defects in the switching layers.
  • FIG. 26 is a diagram illustrating the use of ion implantation or other radiation to create paths or channels of defects in the switching layer structures, according to some embodiments.
  • Beam 150 can be ions such as Au, Pb, U, Er, Yt, Zn, O, Pt, Fe, Bi, H, He, Ga, Hf, N, C, V, Ti, Ta, Ar, Xe, positrons, and electrons.
  • the implantation can take place into and/or through the switching layer 130, either with the top electrode 140 in place, in some embodiments, or without a top electrode yet formed in other embodiments.
  • the implantation disrupts and/or breaks the chemical bonds between the metal and oxygen to form defective metal- oxygen bonds and/or vacancies.
  • transition metal oxide switching materials include TiOx, HfOx, TaOx, VOx, WOx, and ZnOx.
  • Other switching layer materials are also possible, such as C, S, F, and N. According to some embodiments.
  • energy can range from sub KeV to a few thousand KeV, and even to MeV. Dose can range from less than 1 x10 3 /cm 2 to greater than 1 x10 20 /cm 2 .
  • the implantation may be blanket ion implantation or selective area ion implantation.
  • FIG. 27 is a diagram illustrating a forming process as is known in the prior art.
  • FIG. 27 is adapted from "Progress on RRAM as a future non volatile memory” (NVM), P.D. Kirsh, October 28, 201 1 , Sematech.
  • NVM non volatile memory
  • FIG. 27 shows that in known RRAM manufacturing techniques, a forming process is used before the RRAM can exhibit memory effects such as high and low resistance states.
  • the forming process comprises applying a large voltage across the top and bottom electrode, causing a breakdown to occur where bonds between the metal ions and oxygen ions break. This process results in a haphazard path for vacancies, ions and/or electrons to drift under the applied field.
  • ion implantation or other radiation is used to break the bonds. This leads to a more uniform defect path/channel, which results in a reproducible and more manufacturable and reliable RRAM.
  • the ions cause defects such as bond breakage.
  • the defects become the pathways or channels for vacancies, ion and/or electrons to migrate, the combination of which is referred to as a "filament.”
  • heat treatment to activate certain ions can be used.
  • Heat treatment at certain temperatures can be used for partial activation in some embodiments.
  • partial annealing and or heat treatment can be used to remove some, but not all, of the defects.
  • Heat treatment can be used to optimize the resistivity of the RRAM, activate the diode characteristics, and optimize the nonlinearity of the current voltage characteristics. Although some amount of damage or defects should remain in order to have a filament in the RRAM, heat treatment can be used to optimize the characteristics of the filament.
  • Annealing and/or heat treatment temperatures can range from less than 200° C to over 1800° C for a duration from milliseconds to many hours.
  • the anneal can take place in an inert environment such as Ar, or in a reactive environment such as O, or H.
  • the annealing process can have a combination of gases such as N and H, Ar and H, Ar and N.
  • the annealing process can also be in a plasma environment where a gas discharge such as N, O, H or combination of gases can be used in the glow discharge.
  • FIG. 28 is a diagram illustrating the variability of switching parameters, according to known RRAM forming processes.
  • FIG. 28 is adapted from "A Two- Step Set Operation for Highly Uniform Resistive Switching ReRAM by Controllable Filament," Sangheon Lee et al, Sept. 18, 2013, ESSDERC 2013 in Bulgaria.
  • the switching parameters of the RRAM often exhibit variability in Vset 2810, the voltage where the RRAM goes from HRS to LRS, the variability of the value of the LRS 2820, the slope which is the resistance, the variability of the HRS slope 2822 and the variability of the Vreset 2812 which is the voltage to go from LRS to HRS.
  • Vset 2810 the voltage where the RRAM goes from HRS to LRS
  • the variability of the value of the LRS 2820 the slope which is the resistance
  • the variability of the HRS slope 2822 the variability of the Vreset 2812 which is the voltage to go from LRS to HRS.
  • the variability of the switching parameters can be greatly reduced.
  • FIGs. 29A and 29B are diagrams illustrating aspect of RRAM
  • UBL unintentional barrier layer
  • these UBLs 2910 and 2920 are disrupted by ion implantation 2950 through the TE 140 and into the RRAM device.
  • implantation can be to the BE 120, and/or the substrate 1 10.
  • the disrupted UBLs are shown by the dotted dark line 2912 and 2922 in FIG. 29B.
  • the disruption of the UBLs allow the flow of current and the formation of the filaments for RRAM operation where vacancies, ions and/or electrons can migrate, diffuse, drift, hop along the defect paths formed by ion implantation.
  • the circles represent defects, such as defect 2900. Note that prior to ion implantation as shown in FIG. 29A, some native defects exist. After ion implantation the defect concentration increases, as shown in FIG. 29B.
  • the circles can also represent ions implanted such as metal and non-metal ions.
  • FIGs. 30A and 30B are diagrams illustrating forming a thin TE for ion implantation followed by forming a thicker TE, according to some embodiments.
  • a thin TE 3040 can be used for ion implantation through the RRAM structure to reduce the energy that the ions need to penetrate through the TE 3040, SL(s) 130, BE 120 and/or substrate 1 10.
  • the thinned TE 3040 can have a thickness ranging from less than 10 nm to over 100 nm.
  • the UBLs that forms between the BE 120 and SL 130 (UBL 2910 in FIG. 29A) and between SL 130 and TE 3040 (UBL 2920 in FIG. 29A) may be formed as described above.
  • the UBLs are then disrupted by ion implantation 2950 through the thin TE 3040 and into the SL 130, BE 120 and/or substrate 1 10.
  • the disrupted UBLs 2922 and 2912 are shown.
  • the UBLs which act as current barriers, are disrupted such that their chemical bonds are broken by ion implantation into those regions and/or ions passing through those regions. Metal, oxygen, non-metal ions can be implanted or traverse through the interface.
  • ions in thin TE 3040 and SL 130 can be knocked into the contaminated UBL.
  • the disrupted UBLs 2912 and 2922 can than more easily allow current flow.
  • the TE 140 can be thickened by adding material 3042 to reduce series resistance.
  • FIGs. 31 A and 31 B are diagrams illustrating a RRAM structure with top and bottom electrodes made of a combination of metals and semiconductors, according to some embodiments.
  • the TE both thin portion 3040 and the post- implant thickened TE 140
  • BE 120 can be made of a combination of metals and semiconductors, such as silicides (NiSi, PtSi), nitrides (TiN), polycrystalline silicon, highly doped amorphous silicon, and metal layers such as Ti/Pt, Ta/Pt/Au, Ni/Ti and Pt/Ni/Pt.
  • metal 3142 such as Al for example, to reduce series resistance as shown in FIG. 31 B.
  • ion implantation and/or irradiation create defects that disrupt the interfacial UBLs and allow current to pass more easily.
  • Ion implantation can further introduce both metal and non-metal ions into the unintentional interfacial UBLs and/or knock ions from the TE and SL into the UBLs, disrupting them and allow current to more easily flow.
  • the RRAM is manufactured without the prior art forming voltages and accompanying high current that can damage the RRAM and reduce its reliability.
  • U.S. Pat. Nos. 8,062,918, 8,420,478, 8,465,996, 8,872,151 discuss implanting ions partially into the SL to increase defect concentration.
  • those references propose implanting ions before the TE is formed which allows a UBL to form at the interface between the SL and TE.
  • the UBL thickness can be a large fraction of the switching layer.
  • the switching layer can have a thickness as low as 2 nm, while a UBL can have a thickness in the range of 1 -20 nm.
  • a UBL can drastically change the characteristics of the switching layer with its thickness and composition.
  • the UBL can be considered a contamination. This necessitates a further forming process that decreases the reliability of the RRAM and increases its variability.
  • the references teach implanting the ions/defects such that they remain
  • U.S. Pat. No. 8,809,159 discusses a thin TE of 20 nm that is transparent to radiation that can be formed prior to irradiation.
  • the transparent TE is required since very low energy is used for irradiation and need to irradiate through the TE and into the SL. This low energy of tailoring the irradiation primarily in the SL results in non-uniform distribution of defects and may not disrupt a UBL formed at the SL and BE interface.
  • ion implantation/irradiation into and/or through the unintentional barrier layer(s) is performed using ions with multiple implant energies, each being tailored for the TE, UBLs, SL, BE, and/or through the UBLs and into the BE and/or substrate. No further critical interfaces that can unintentionally impede current flow are formed after the ion implantation process and irradiation process.
  • multiple energies of implanted ions range from Kev to over MeV.
  • protons can be implanted to 2 microns at 200 KeV that for a single layer RRAM crossbar architecture puts the bulk of the protons into the substrate.
  • the transit of the protons through the TE, SL and BE and into the substrate create a uniform distribution of defects channel/paths for the vacancies, ions, and/or electrons to migrate and form a filament.
  • Other ions besides protons can be used, including metal and/or non-metal ions. Greater uniformity of performance across an entire chip that includes larger numbers of RRAM cells can be achieved by uniformly creating enhanced channels in the switching layers through the use of radiation and/or ion implant damage.
  • FIG. 32 is a flow chart illustrating processing blocks for a basic RRAM, according to some embodiments. Processing steps for a RRAM device / crossbar are shown starting with block 3210 where the substrate may contain CMOS integrated circuits of active and possibly passive devices to interface with the RRAM to other information processing units such as CPUs.
  • the bottom electrode is formed, followed in block 3214 in which switching layer(s) are formed. According to some embodiments, the switching layer can also include resistors, tunnel and/or diode layers.
  • the top electrode is formed. Ion implantation and/or irradiation is (are) performed through the entire RRAM structure in block 3218.
  • the implantation or irradiation of block 3218 can be through the entire stacked RRAM structure with sufficient energy to penetrate the RRAM and/or the entire RRAM stack. According to some other embodiments, intermediate implantation / irradiation steps can be carried out following the formation of one or more stack layers.
  • the top electrode is thickened, if desirable.
  • the ion implant and/or irradiation of block 3218 may consist of multiple energies, doses and different ions, electrons, and protons. According to some embodiments, a uniform distribution of defect channels extends from the top electrode, through the switching layer(s) and into the bottom electrodes and/or into the substrate.
  • the entire silicon wafer is irradiated and/or ion implanted which is conducive to high volume silicon wafer processing.
  • the device is irradiated in situ. That is, irradiation is performed after the deposition of the BE, the SL and the TE, for example as shown in FIG. 32.
  • high dose from 100 rad to 10 million rad in silicon
  • electromagnetic radiation such as X-rays in the range of about 3 KeV to about 100 KeV, is used to create a more uniform channel of defects through the SL.
  • One rad in silicon, or 1 rad (Si) is generally defined as 100 ergs energy absorbed per gram of silicon.
  • x-ray radiation with photon energies ranging from 3 Kev to 100 Kev is used.
  • photon energies ranging from 3 Kev to 100 Kev.
  • U.S. Patent 8,809,159 discusses the use of very low radiation photon energies (wavelengths of 200-600 nm or 6.2 to 2.1 ev), resulting in extremely narrow range of depths of defect generation in the SL.
  • This low-energy approach also necessitates that the irradiation is preformed before the top electrode is deposited or using a very thin layer of top electrode (10 to 20 nm in thickness).
  • FIGs. 33A-D are plots illustrating simulated result of Hf ions implanted according to a known technique.
  • U.S. Pat. No. 8,872, 151 discusses implanting Hf ions into a 30nm HfOx switching layer without a top electrode at 50KeV and 22.5KeV.
  • the plots 3310, 3312, 3314 and 3316 are Monte Carlo simulations of these conditions.
  • the vacancies generated within the switching layer are very non-uniform. This non-uniformity with areas devoid of vacancies can cause inconsistent and irreproducible RRAM switching characteristics and requires further forming to initiate switching, as volumes without vacancies will require forming to generate vacancies.
  • Forming may also be required to break the UBLs between the HfOx and the bottom electrode, and between the top electrode and the HfOx that is deposited after the ion implantation process.
  • Non-uniformity of the vacancy concentration is over 95% within the switching layer. This non-uniformity can cause constrictions in the vacancy concentration resulting in mobility degradation that can eventually lead to RRAM failure after many cycles between HRS (high resistance state) and LRS (low resistance state). Lower concentration of vacancies can also lead to low vacancy mobility and lower conductivity.
  • FIG. 34 is a diagram illustrating relative thicknesses of a simple RRAM structure of the electrodes and the switching layers, according to some
  • Implantation process 3450 can be performed using a variety of ion species, a variety of implant energies and a variety of doses. Several implant scenarios into the structure shown in FIG. 34 were simulated and the results are illustrated in FIGs. 35A-D, 36A-D, 37A-D and 38A-D. It has been found that providing a relatively deep implant process, such that the bulk of ions are implanted well into the substrate 1 10, results in greatly increased uniformity of the defect profile.
  • the vacancy concentration uniformity within the switching layer in this case is better than 10%.
  • the bulk of the ions are implanted outside the switching layer, which can provide acceptable vacancy concentration uniformity.
  • 90% or more of the ions are implanted outside the switching layer.
  • 70% or more of the ions are implanted outside the switching layer.
  • 50% or more of the ions are implanted outside the switching layer.
  • 30 % or more of the ions are implanted outside the switching layer.
  • 10% or more of the ions are implanted outside the switching layer.
  • Uniformity of the vacancy concentration within the switching layer can range from better than 5% to better than 80% in variations about a mean value.
  • the vacancy concentration uniformity can be better than 5% in variation. In some cases it can be better than 10%. In some cases, it can be better than 20%. In some cases, the vacancy concentration uniformity can be better than 30%. In some cases it can be better than 40%-50%. In some cases, the vacancy uniformity can be better than 60-80% in variation about a mean value.
  • multiple ion implants with different ion species, energies and/or doses are used to improve the RRAM reproducibility, reduce variability and improve reliability.
  • a combination of shallow and deep ion implants are used.
  • a first shallower ion implantation effectively disrupts the UBLs between the TE and SL and/or SL and BE layers. This is followed by a second, deeper implant that is effective in generating a uniform defect profile across the entire SL layer for.
  • Other combinations of ion implantation can also be incorporated into the RRAM manufacturing process, for example implants that convert transition metals to transition metal oxides in situ (nitrides, carbides, fluorides, chlorides, oxides and any combinations thereof are other examples).
  • FIGs. 35A-35D are plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 34 at 600 KeV.
  • the Ar ions can be singly ionized and accelerated at 600KeV or doubly ionized and accelerated at 300 KeV.
  • the Ar ion peak is just less than 500 nm into the Si02 layer.
  • Plot 3512 is an expanded plot of the TE 140, SL130 and the upper part of BE 120.
  • Plots 3514 and 3516 are collision events plots that show the vacancies generated by the implantation.
  • FIGs. 36A-D are plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 34 at 800 KeV.
  • the Ar ion peak is deeper than 500 nm into the Si02 layer.
  • Plot 3612 is an expanded plot of the TE 140, SL130 and the upper part of BE 120.
  • Plots 3614 and 3616 are collision events plots that show the vacancies generated by the implantation.
  • the expanded plot 3616 shows that vacancies are generated approximately uniformly within the switching layer 130 with better than 50% uniformity.
  • the uniformity of the vacancy concentration in the switch layer is better than 3%.
  • Ar is an inert ion used primarily to generate vacancies and defects, and to knock electrode ions into the interfaces to disrupt the UBLs that can form during processing which as cause the RRAM device to require a breakdown-level forming process.
  • FIGs. 37A-D are plots illustrating simulation results of Xe ions implanted into a structure as shown in FIG. 34 at 1600 KeV, or triply ionized and accelerated at 533 KeV. From plot 3710 it can be seen that the ion distribution peak is approximately 500 nm from the surface.
  • Plot 3712 is an expanded plot of the TE 140, SL130 and the upper part of BE 120. From plots 3714 and 3716 it can be seen that the vacancies/defects generated within the switching layer 130 and at the TE/SL and SL/BE interfaces are approximately uniform with variation in the vacancy concentration of less than 10%.
  • the uniformity is better than 2%.
  • the approximate uniform distribution and/or distributions without any voids within the switching layer and/or at the interfaces of vacancies/defects within the switching layer and at the interfaces can result in low variability RRAM switching characteristics and improve reproducibility and reliability.
  • 1600 KeV of acceleration energy is used in this example, other energies less than or greater than 1600 KeV can also be used.
  • multipli-ionized ions can use a correspondingly lower accelerating voltage, and achieve similar depth of implant.
  • E (energy) ZeV can be used where Z is the ion charge number, V is the accelerating voltage and e is the electron charge.
  • a triply ionized Ar ions having a 600 KV accelerating voltage will have an energy of 1800 KeV.
  • energies where the ion peak lies within the bottom electrode can also be used; consideration is given to any CMOS circuits and/or devices that may be beneath the RRAM and the energies can be tailored to minimize ion implantation into the CMOS.
  • FIGs. 38A-D are plots illustrating simulation results of Bi ions implanted into an example structure as shown in FIG. 34 at 3200 KeV. From plot 3810 it can be seen that Bi ions are implanted to a depth of approximately 500 nm.
  • Plot 3812 is an expanded plot of the TE 140, SL130 and the upper part of BE 120. Vacancies within the switching layer 130 and at the interfaces are approximately uniform as can be seen in plots 3814 and 3816. Bi ions can be multiply ionized to reduce the acceleration voltage. In addition, energies other than 3200 KeV may be used.
  • Bi ions can be implanted into the bottom electrode and/or top electrode interface, within in the switching layer, and/or at the bottom electrode interface and followed by higher energy Bi and/or other ions implant to generate vacancies without voids and/or approximately uniform vacancies within the switching layer(s) and the interfaces.
  • the uniformity of the vacancy concentration is approximately less than 5%. Not considering portions than might be due to simulation artifacts, better than 2% in vacancy concentration uniformity is achieved.
  • Higher energy ion implantation can also have less variation to uniformity of vacancies, defects, ion distributions due to variations in thicknesses of the RRAM devices that can be caused by processing for example.
  • FIG. 39 is a log-log plot illustrating the relationship between X-ray linear absorption coefficient in Silicon (in cm -1 ) and X-ray photon energy (in Kev), for energies between 1 Kev and 100 Kev.
  • the use of high dose electromagnetic irradiation, such as X-rays, are shown to facilitate RRAM manufacture, in the following examples.
  • the linear absorption coefficient in silicon is shown broken into two curves 3910 and 3912 for clarity.
  • ABS NIST X-ray mass attenuation coefficient for Silicon.
  • ⁇ is in the range of 1000 to 0.4 /cm (or cm "1 ).
  • T is the silicon thickness in cm.
  • the ABS is virtually linearly proportional to the RRAM device thickness. A change of several percent in the RRAM thickness changes the ABS by only several percent.
  • this irradiation in the range of 26 KeV, or 20-30 KeV, is very insensitive to thickness variations and is particularly well suited to in-situ irradiation of stacked cross bar RRAM devices where there is little tolerance for variations.
  • An example of a stacked structure of three RRAM devices is shown in FIG. 45, can be readily irradiated in-situ using the ranges described above.
  • in-situ irradiation of RRAM structures having larger numbers of stacked devices is also practical using an energy range of 20-30 KeV.
  • stacks of 5 to 15 or more stacked RAMM devices are irradiated in-situ using an energy range of 20-30 Kev.
  • Vertical RRAM structures an example of which is shown in FIG. 53, are also well suited for in-situ irradiation using electromagnetic energy, especially in the range of 20-30 KeV
  • x-rays can be provided with
  • Bremsstrahlung devices such as those used for medical purposes
  • electron synchrotron devices since both electrodes are formed prior to the in- situ irradiation of the switching layer, the irradiation can disrupt any UBLs formed unintentionally via contamination during processing.
  • CMOS may be beneath and/or around the RRAM crossbar
  • precautions can be taken to mask those sensitive regions to exposure to radiation by a thin layer, e.g. several microns thick, of heavy metal such as W, Ta, Bi, Au and Pb.
  • a thin layer e.g. several microns thick, of heavy metal such as W, Ta, Bi, Au and Pb.
  • W, Ta, Bi, Au, Pb and/or other metals is first deposited to shield other components that are underneath or around this location. This suggests that impractically thick shielding would be required for x-ray photon energies much beyond 60 KeV. This effect was not appreciated in prior art proposals such as U.S. Pat. No.
  • FIG. 40 is a diagram illustrating a simple cross section of a RRAM structure that has the switching layer(s) normal or nearly normal to the substrate, according to some embodiments.
  • RRAM switching layers can be horizontal (parallel to the substrate and/or almost horizontal) and/or vertical (normal to the substrate and/or almost vertical) and any angle from normal to parallel to the substrate.
  • Ion implantation and/or irradiation with X-ray/Gamma rays can impinge the RRAM structure at angles ranging from normal to the substrate to off normal, for example from right angle to an acute or obtuse angle to the substrate.
  • radiation 4050 e.g.
  • ion implantation and/or radiation is directed at the RRAM structure at an off normal angle, for example approximately 45 degrees through the top electrode 140 and into the switching layers 4030, 4032 and 4034, into the bottom electrode 120 and/or into the substrate 1 10.
  • the ion implantation and/or irradiation 4050 can also be normal or almost normal to the substrate 1 10 and cause vacancies/defects within the switching layer(s) and interface(s).
  • the angle of the ion implant or other radiation can range from less than 5 to 80 degrees or more off normal.
  • FIG. 41 is a diagram illustrating a cross section of a RRAM structure where the ion implant and/or other radiation is performed normal or almost normal to a primary planar surface of the substrate, according to some embodiments.
  • the RRAM structure is as shown in FIG. 40.
  • the ion implantation and/or irradiation by X-ray/Gamma rays 4150 can impinge at an angle that is normal and/or almost normal to the primary planar surface of the substrate.
  • the energies are chosen such that the ions and/or radiation generate vacancies/defects through the electrodes and into the switching layer(s) and interfaces, into the bottom electrode and/or into the substrate.
  • ions and/or radiation penetrate the switching layers 4030, 4032 and 4034 edge wise after the interfaces between switching layers and any other layers for current control and the electrodes, such as the bottom electrode 120 and top electrode 140, are fabricated.
  • the ions and/or radiation 4450 can disrupt any UBLs formed unintentionally via contamination during processing. This disruption improves RRAM switching characteristics without relying on a forming process that applies a near breakdown voltage-current to the devices. Vacancies and defects are generated in the switching layers 4030, 4032 and 4034.
  • ions such as O, transition metal ions, other non-transitional metal ions, inert ions such as Ar, Xe, non-inert ions such as H, CI, N, F, and other ions such as Li, Na, K, C to name a few can also be implanted into the switching layers edge wise and/or almost edge wise.
  • the distribution of ions and/or vacancies/defects is along the direction of the impinging ions.
  • the width of the switching layer is greater than the thickness of the switching layer, which ranges from about 2 to about 40 nm.
  • a distribution of vacancies and/or ions can therefore be placed along the width of the switching layer(s) rather than the depth of the switching layer(s). This technique can advantageously localize the vacancies/ions within the width of the switching layer(s) that can further confine the switching region within the switching layer(s) to improve reproducibility and reliability.
  • FIG. 42 is a diagram illustrating a stacked RRAM structure with two or more stacks where the plane of the switching layers are normal and/or almost normal to the surface of the substrate, according to some embodiments.
  • energies for ion implant and/or radiation such as X-ray/Gamma rays are selected such that the ions and/or radiation penetrate the entire stack 4200 through the electrodes 4240, 4220, 140 and 150, into the switching layers such as 4230 and 4232, and into the substrate 1 10.
  • the ions and/or radiation also penetrate through the interfaces between electrodes and switching layers where UBLs may exist thereby disrupting the UBLs.
  • the angle of the impinging ions and/or irradiation can be normal and/or almost normal to the substrate such as in example implant 4252 so that the switching layer(s) are penetrated edge wise. According to some other embodiments, the impinging ions and/or radiation can be off normal as in example implant 4250.
  • FIG. 43 is a diagram illustrating a cross section of a basic simplified RRAM structure with vertical and/or almost vertical switching layer(s) with ion implantation and/or radiation impinging edgewise into the switching layers, according to some embodiments.
  • the edgewise direction allows for localization of the ions/vacancies/defects in a region along the width of the switching layers rather than along the depth or thickness of the switching layer.
  • the vacancy concentration is approximately uniform along the direction of the current flow.
  • An example ion and/or defect profile is shown in inset diagram 4360. Ions with high enough energies can also penetrate along the entire width of the switching layers so as to implant into the bottom electrode 120 and/or substrate 1 10.
  • multiple ion species, energies and/or doses can be used to optimize the RRAM switching characteristics, reproducibility and reliability, as well as to reduce RRAM variability.
  • radiation such as X-rays and Gamma rays can also be used together with ion implantation and/or separately, to generate defects along the entire width of the switching layers.
  • the ion implantation and/or radiation into the edge of the switching layers is performed after the electrodes and/or non-switching layers (such as current control layers, non-linear diode and/or tunnel layers) are in contact with the switching layers. This can be beneficial, for example, for disruption of any UBLs formed as interfaces.
  • the ion implant and/or radiation can be initiated when the switch layers are deposited on the bottom electrode prior to the top electrode deposition.
  • the switching layers are typically a lower density material than the electrodes, ion implantation and/or radiation into the edge of the switching layer(s) may require less energy overall, while still achieving localized or non-localized uniformity, depending on the energy distribution of ions and/or vacancies.
  • multiple ion species, energies, and/or doses may be used to optimize RRAM characteristics.
  • the lower energy for ion implantation and/or radiation also minimizes unintentional effects of CMOS circuits that may be beneath and/or around the RRAM crossbar and reduces any shielding
  • the edge ion implant and/or radiation also acts to disrupt the UBLs that can form due to processing which would otherwise negatively impact RRAM characteristics.
  • the disruption of the UBLs can allow a better contact between electrode and switching and/or non-switching layers.
  • FIG. 44 is a diagram illustrating a simple RRAM structure where the switching layer includes ZnO nanorods, according to some embodiments.
  • the ZnO nanorods can be core shell, with other switching layers such as HfOx and/or with other current control layers such as resistive layers, diode layers and/or tunnel layers.
  • the ZnO nanorods are shown in the switching layer 4430 between a bottom electrode 4420 and top electrode 4440.
  • BE 4420 is aluminum doped ZnO (AZO)
  • TE 4440 is Au.
  • Implantation process 4450 can include hydrogen ions implanted through the top electrode 4440, into the ZnO nanorods, into the bottom electrode 4420 and/or into substrate 1 10.
  • the H ions can be used to create O vacancies in the ZnO nanorods. Doping the ZnO nanorods with H ions, with or without thermal anneal/ heat treatment, either uniformly and/or non-uniformly, can adjust the switching characteristics of the RRAM. In addition, by implanting through the top electrode, any unintentional barrier layer (UBL) at the interfaces will be disrupted so as to provide good electrical characteristics. According to some embodiments, a higher energy ion implant can also be used to disrupt any UBL at the interface between the nanorods and the bottom electrode.
  • UBL unintentional barrier layer
  • other ions may also be used in the ion implantation, such as Zn, O, Ar, Xe, Al, Au, Ti, Hf, to dope and/or to create defects and/or vacancies within and along the nanorod interfaces with other nanorods.
  • multiple energies, doses and/or species can be used to optimize the switching characteristics of the RRAM.
  • TE 140 is Pt having 50 nm thickness
  • the SL 130 is hafnium oxide having 20 nm thickness
  • the BE 120 is Ti or W having 100 nm thickness
  • the substrate 1 10 is silicon dioxide having 2000- 4000 nm thickness on Si wafer.
  • Alternative examples include substituting titanium oxide, zirconium oxide, tungsten oxide, tantalum oxide, vanadium oxide or a combination of such oxides for the SL 130.
  • the electrodes TE and BE can vary in thickness and can be made of TiN, AIN, silicide, W, Ni, Au, Pt, Cr, V, Ta, Nb, Al, Cu or a combination of these. Following are specific implant schedules that can be used with the above structure(s):
  • a hydrogen implant can be added to any of the above example implant schedules 1-3 and 5-14. Additionally, the energy and dose of any of the above implant schedules can be varied by +/- 50%.
  • FIG.45 is a diagram showing a structure of multi-layer RRAM devices according to some embodiments.
  • the structure 4500 shown in FIG.45 is an example of three vertically stacked RRAM devices. According to some
  • the TEs 4540, 4542 and 4544 are made of 100 nm of Pt; switching layers (SLs) 4530, 4532 and 4534 are made of 20 nm of TiOx; and Bes 4520, 4522 and 4524 are made of 100 nm Ti.
  • the multi-layer structure 4500 consists of 3 RRAM devices on S1O24510 that can be formed on Si, for example.
  • Ion implantation 4550 is performed.
  • Ar ions are implanted at 800 and/or 1600 KeV.
  • H ions are implanted at 100 and/or 200 KeV into the structure 4500.
  • Ar ions are inert and do not react with any material.
  • H can be reactive and the bulk of the H ions are deposited into the S1O2 layer 4510 at 200 KeV. Whereas, the bulk of the Ar ions at 800 and 1600 KeV reside in the second and third RRAM device respectively.
  • FIGs. 46A-46H are plots illustrating simulation results of Ar and H ions implanted into a structure as shown in FIG. 45 at various energies, according to some embodiments. Plots are shown of the range of the ions and the collision events.
  • the doping concentration of ions/cm 3 can be obtained by multiplying dose (ions/cm 2 ) with the vertical axis of (atomes/cm 3 )/(atomes/cm 2 ).
  • the collision events plot FIGs.
  • vacancies concentration can be approximated by multiplying the (dose) (r) (10 8 ) with the vertical axis of number/(angstrom-ion) where "r" is the number of ions to create one vacancy.
  • “r” is the number of ions to create one vacancy.
  • FIGs. 46A and 46B it can be seen that Ar at 800 KeV can only penetrates 2 RRAM devices.
  • FIGs. 46C and 46D it can be seen that Ar at 1600 KeV Ar penetrates all 3 RRAM devices.
  • FIGs. 46E-46H it can be seen that H at 100 and 200 KeV is able to penetrate all 3 RRAM devices.
  • FIGs. 47A-47B are expanded plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 45 at 800 KeV, according to some embodiments.
  • Ar 800 KeV vacancies are generated in the switching layer of the first two RRAMs.
  • the first RRAM switching layer 4534 has better than 20% uniformity in the vacancy concentration and the second RRAM switching layer 4532 has better than 10% vacancy concentration uniformity.
  • the device to device vacancy concentration variation is better than 10%.
  • FIGs. 48A-48C are expanded plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 45 at 1600 KeV, according to some embodiments. It can be seen that the switching layers 4534, 4532 and 4530 for RRAM1 , RRAM2 and RRAM3 respectively all have vacancy concentration variations better than 20%. Ignoring likely simulation artifacts near the boundary of the switching layers, the uniformity of the vacancy concentration is better than 5% within each RRAM's switching layers. The device-to-device variation in the vacancy concentration better than 15%.
  • FIGs. 49A-49C are expanded plots illustrating simulation results of H ions implanted into a structure as shown in FIG. 45 at 200 KeV, according to some embodiments.
  • the vacancy concentration uniformity is better than 60% all 3 RRAM SLs (4534, 4532 and 4530) and device-to-device uniformity is
  • FIG. 50 is a diagram showing a structure of multilayer of RRAM devices according to some embodiments.
  • the structure 5000 shown in FIG. 50 is an example of three vertically stacked RRAM devices. According to some
  • the TEs 5044, 5042 and 5040 are made of 100 nm of Al; SLs 5034, 5032 and 5030 are made of 20 nm of TiOx 20 nm; and BEs 5042, 5040 and 5020 are made of 100 nm of Al.
  • electrode layers 5040 and 5042 serve as both TE and BE functions for different RRAM devices.
  • a 3000 nm S1O2 underlies the RRAM devices.
  • the structure is implanted with Ar ions using energies of 600 and 800 KeV. H ions can also be implanted using energies of 100 and 200 KeV.
  • the stacked RRAM structure of FIG. 50 only shows the most basic structures for simplicity. Other layers can be included and implant energies of ion such as Ar, H, Xe, Ti, Hf can be adjusted accordingly to increase uniformity of distribution of ions and vacancies.
  • FIGs. 51A-51 E are plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 50 at 600 KeV, according to some embodiments. It can be seen in FIGs. 51A-51 E that better than 10% vacancy concentration variation within each device can be achieved. Ignoring likely simulation interface artifacts the vacancy concentration is better than 5%. Device- to-device variation is better than 20% with Ar ion implant at 600 KeV.
  • FIGs. 52A-52E are plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 50 at 800 KeV, according to some embodiments. It can be seen in FIGs. 51 A-51 E that better than 20% vacancy concentration variations within a device and deviceto-device can be achieved using a 800 KeV Ar ion implant into the 3 RRAM stack with composition as shown in Fig. 50. In some cases, better than 15% in vacancy concentration variation can be achieved with likely simulation artifacts removed.
  • FIG. 53 is a diagram showing a structure of vertically stacked RRAM devices according to some embodiments.
  • the structure 5300 shown in FIG. 53 is an example of a vertical RRAM stack (VRRAM) of 5 RRAM devices having a vertical electrode 5320; horizontal electrodes 5340, 5342, 5344, 5346 and 5448; and a switching layer 5330 disposed between the horizontal and vertical electrodes. As shown in the FIG. 53, the switching layer 5330 is approximately normal to the substrate 5310. Between the horizontal electrodes 5340, 5340, 5342, 5344, 5346 and 5348 are electrically insulating spacer layer 5362, 5364, 5366 and 5368. Additionally, insulation layer 5314 is provided below horizontal electrode 5348, and insulation layer 5360 is provided above horizontal electrode 5340.
  • VRRAM vertical RRAM stack
  • CMOS electronics can exist in substrate 5310 beneath the VRRAM and a layer 5312 made of a material such as W (tungsten), Mo (molybdenum), Hf and/or Ta, can be deposited between the CMOS electronics and the WRAM to stop ions and/or radiation from penetrating into the CMOS layers.
  • the thickness of the electrodes 5340, 5340, 5342, 5344, 5346 and 5348 can range from 50 to 200 nm and the insulating spacer layers can range from 50 to 200 nm.
  • the heavy metal stopping layer 5312 can range from 100 to 2000 nm.
  • the total stack of VRRAM can range from 550 nm to 2200 nm approximately. In some cases it can be 4500 nm approximately.
  • Ion implantation 5350 approximately normal to the surface that can penetrate into the edge of the switching layer either at a normal and/or off-normal angle to generate uniform defects along the entire length of the switch layer.
  • Ion implantation and/or radiation can be implemented when the vertical electrodes and the horizontal electrodes are in place but prior the adding the vertical connecting wires. In some cases, ion implantation and or radiation can be implemented after the connecting wires for the vertical electrodes are in place. In some other cases, the ion implantation and/or radiation can be implemented after the entire VRRAM structure is formed. In some other cases, some masking for selective area ion implantation and/or radiation can be implemented.
  • Mask material for selective are ion implantation and/or radiation can be a metal such as Hf, Ta, W, Pb, or Au.
  • selective area ion implantation and/or radiation may be performed on certain areas of the switching layer, such as areas that are directly between the vertical and horizontal electrodes.
  • Heavy metals can be used shielding the CMOS in a selective area ion implant and/or radiation.
  • Other thick materials such as polyimide can also be used according to some embodiments.
  • FIG. 54 is a diagram showing structure of vertically stacked RRAM devices that is simplified for purposes of simulation, according to some
  • the VRRAM structure 5400 comprises switching layer 5330 of HfOx at 2500 nm thickness, followed by insulation layer 5314 made of 2000 nm of S1O2, layer 5312 made of 2000 nm of W, and Si layer 5310 of 10000 nm thickness. Ions are implanted to a depth of approximately 4500 nm to maximize uniformity of the vacancy concentration distribution within the HfOx switching layer 5330.
  • the W layer 5312 is used to stop ions from penetrating into the Si layer 5310 which may contain CMOS electronics for controlling the VRRAM and other memory functions.
  • FIGs. 55A-55B are plots illustrating simulation results of Ar ions implanted into a structure as shown in FIG. 54 at 17.5 MeV, according to some embodiments. The plots show that the Ar ions penetrate to a depth of
  • FIGs. 56A-56B are plots illustrating simulation results of H ions implanted into a structure as shown in FIG. 54 at 540 KeV, according to some embodiments. The plots show that the Ar ions penetrate to a depth of
  • the W layer 5312 is effective in stopping ions from penetrating into the Si layer which may contain CMOS circuits and electronics.
  • FIG. 57 is a diagram showing structure of vertically stacked RRAM devices that is simplified for purposes of simulation, according to some other embodiments.
  • the VRRAM structure 5700 includes switching layer 5730 made of 2500 nm of TiOx.
  • Insulation layer 5314 is 2000 nm of S1O2
  • layer 5312 is 2000 nm of W
  • layer 5710 is 10000 nm of Si.
  • the W layer 5312 or similar heavy metal layer is used to stop ions from penetrating into the Si layer 5710 which may contain CMOS circuits and electronics.
  • the W layer 5312 thickness can range from 500 to 2000 nm approximately depending on ion energies.
  • the simulation conditions included Ar ions at 17 MeV and H ions at 520 KeV, which resulted in an ion range of approximately 4500 nm.
  • the ion range peak was placed outside of the switching layer 5730 to maximize uniformity of the vacancy concentration distribution along the length of the switching layer such that the VRRAMs structure has relatively low variability between the RRAMs at different levels.
  • FIGs. 58A-58D are plots illustrating simulation results of Ar and H ions implanted into a structure as shown in FIG. 57 at 17 MeV and 520 KeV
  • concentration distribution along the length of the TiOx switching layer 5730 is better than 40% variation in uniformity for the Ar implant and approximately 30% variation in uniformity for the H implant energy with the likely simulation artifacts removed.
  • the W layer 5312 effectively stops ions from penetrating into the Si layer 5710 which may contain CMOS circuits and electronics for controlling the VRRAM memory stack and other memory functions.
  • FIG. 59 is a diagram showing structure for RRAM devices this was used to investigate sensitivity of vacancy concentration in switching layers as a function of top electrode thickness variation due to manufacturing.
  • the structure 5900 includes a top electrode 9540 made of Pt which had a variable thickness. For the study, the thickness was varied by 20% from 80, 100, 120 nm.
  • the switching layer 5930 is made of 20 nm of TiOx
  • the bottom electrode 5920 is made of 100 nm of Ti
  • the S1O2 layer 5910 is 2000nm thick.
  • Ar ions were accelerated to an energy of 1600 KeV such that the Ar ion range or peak was outside the switching layer 5930.
  • FIGs. 60A-60C are plots illustrating simulation results of Ar ions implanted at 17 1600 KeV into a structure as shown in FIG. 59 having top electrode 5940 thicknesses of 80, 100 and 120 nm, according to some
  • FIGs. 60A, 60B and 60C show the simulation results for the TE (Pt) layer 5940 thickness of 80 nm, 100 nm and 120 nm, respectively. It was found that the variation of the vacancy concentration in the switching layer (5030) due to Pt top electrode (5940) thickness variation was better than 5%.
  • FIGs. 61 A and 61 B are diagrams illustrating an example of constrictions in vacancy filaments. They are adapted from Chen et al., "Understanding of the Endurance Failure in Scaled Hf02-based 1 T1 R RRAM through Vacancy Mobility Degradation" IEDM12-482, 2012, which is incorporated herein by reference. That non-uniform distribution of oxygen vacancies in the case of transition metal oxide switching layers can result in constrictions in the vacancy filament.
  • Chen et al. “Understanding of the Endurance Failure in Scaled Hf02-based 1 T1 R RRAM through Vacancy Mobility Degradation” IEDM12-482, 2012, which is incorporated herein by reference. That non-uniform distribution of oxygen vacancies in the case of transition metal oxide switching layers can result in constrictions in the vacancy filament.
  • constrictions 6134 and 6138 can result in degradation of the vacancy mobility leading to degradation of the RRAM.
  • the degradation of the vacancy mobility results in higher resistivity and consequently higher joule heating that can further anneal the defect reducing the number of vacancies resulting in the LRS becoming more resistive.
  • the formation of conductive filaments with oxygen vacancies, metallic defects, oxygen ions, neutralized oxygen vacancies is discussed in Syu et al. "Redox Reaction Switching Mechanism in RRAM Device With Pt/CoSiOTiN Structure," IEEE ELECTRON DEVICE LETTERS, VOL. 32, NO. 4, APRIL 201 1 (hereinafter "Syu 201 1 "), which is incorporated herein by reference.
  • Non-uniform distribution in the direction of current flow across the switching layer between the electrodes in horizontal and vertical RRAMs can result in constrictions in the filaments which lead to degradation in vacancy mobility. This degradation in mobility of the vacancies then result in the degradation of the LRS which causes degradation in reliability of the RRAM.
  • the variation of vacancy concentration in the non-uniform distribution of vacancies due to forming process results in variability of all the switching parameters of the RRAM, such as HRS, LRS, Vset and Vreset and V forming.
  • Vread the voltage at which the memory is interrogated weather in the HRS or LRS
  • the current associated with Vread can also vary within a single RRAM and/or other RRAM devices on the memory cell which can contain millions to trillions of RRAM cells. This variability is highly undesirable.
  • the filaments formed by a forming process is haphazard and follow a path connecting local defects such as grain boundaries and defects in deposition of the switching layer. See, e.g. Raghavan et al., "Variability model for forming process in oxygen vacancy modulated high-j based resistive switching memory devices," Microelectronics Reliability 54 (2014) 2266-2271 , which is incorporated by reference herein.
  • the haphazard length of the filament and the concentration of vacancies which can affect mobility of the vacancies and add to the variability of voltages and current of the RRAM, which in turn affects HRS, LRS, Vset, Vreset, Vread, Vforming.
  • the variation in length and mobility of the vacancies can also affect the speed of the device, for example the switching time between the memory states of HRS and LRS.
  • Kinks in the filaments where there may be constrictions in the vacancy concentration can result in lower mobility of the vacancies and therefore longer switching time.
  • mobility of approximately 10 2 cm 2 /V-s with an applied electric field of 1V/cm result in a switching time of 2-20 nS (nanoseconds).
  • the filament can be partially non-conducting, resulting in a HRS.
  • the non-conducting segment can be 2-4 nm within a switching layer thickness of 20 nm.
  • the mobility can be 10 cm 2 /V-s with a switching time of tens of nanoseconds. Mobility of the vacancies can also be approximately 10 "2 cm 2 /V-s or lower which results in microsecond switching times.
  • FIG. 62A is a diagram illustrating a uniform distribution of the vacancies within the switching layer, according to many of the embodiments described herein.
  • the uniform vacancy distribution 6232 of FIG. 62A can be achieved using, for example, the techniques which were simulated and discussed with respect to FIGs. 34, 35A-D, 36A-D, 37A-D and 38A-D.
  • FIG. 62B is a diagram illustrating vacancy distribution resulting from the known technique of ion implanting prior to the formation of the top electrode.
  • the distribution of vacancies 6234 in FIG. 62B can result from known ion implantation techniques such as illustrated FIGs. 33A-D.
  • the vacancies 6234 are non-uniform within the switching layer in the direction of the current path with constrictions that can result in degradation of the vacancy mobility and eventually resulting in the degradation of the reliability of the RRAM.
  • the constriction in the number of vacancies can also result in lower vacancy mobility and therefore can adversely affect the switching time of the RRAM.
  • there are cases in the known ion implantation schemes where the vacancies and/or defects due to ion implant do not extend all the way to the bottom electrode. In such cases, a forming process is necessary to form the filament. From the simulation results illustrated in FIG. 33A- D it can be seen that the vacancy ranges from 0-2.4 in one case.
  • x dose / angstrom for example a dose of 1 x10 12 ions/cm 2 gives a vacancy concentration from 0 to 2.4x10 20 /cm 3 where it is assumed each ion collision give rise to one vacancy. Often this is not the case and every 100 ion collisions result in one vacancy.
  • Vacancy concentration can be in the range of 2x10 7 /cm 3 to 5x10 22 /cm 3 . In certain cases, a range of 5x10 21 /cm 3 to 5x10 22 /cm 3 can be achieved. In another case for lower current, 5x10 20 /cm 3 to 5x10 2 /cm 3 can be achieved. For very low current operation, in the micro amperes and/or sub microamperes range, 5x10 9 /cm 3 to 5x10 20 /cm 3 can be achived. In the high nanoamperes range for ultra-low energy RRAM, vacancy concentration can be in the range 1 x10 7 /cm 3 to 5x10 9 /cm 3 .
  • concentration variations of less than 80% within the switching layer can be achieved. Variability of the RRAM parameters such as HRS, LRS, Vset, Vreset and Vread are reduced since the vacancy concentration and mobility are uniform and have no constrictions. In some cases, the vacancy concentration can be better than 60% in variation within the switching layer, in some cases better than 40%, in some cases better than 20% and in some cases better than 10%. See FIGs. 35A- D, 36A-D, 37A-D and 38A-D for single switching layer ion implants. The deep heavy ion implants produce the most uniform vacancy concentration within the switching layer.
  • Uniform concentration of vacancies in the switching layer can result in uniform vacancy mobility which not only reduces the variability of the RRAM switching parameters such as HRS, LHS, Vset, Vreset, Vread, but also reduces switching time and improves reliability during cycling between the memory states of HRS and LRS.
  • RRAM switching parameters such as HRS, LHS, Vset, Vreset, Vread
  • vacancy concentrations 20% or better
  • uniformity of Vset and Vreset, HRS and LRS, Vread and switching time can be expected to have variability of less than approximately 10% within a single device.
  • RRAM cell-to-cell variations should also approximately be better than 10%.
  • RRAM parameter variability can be better than 5% within a single RRAM and cell-to-cell RRAM. In certain cases an RRAM parameter variability of better than 2% can be achieved.
  • FIGs. 63A-63D are diagrams showing a simplified sequence of processes for manufacturing a portion of RRAM devices, according to some embodiments.
  • ions are implanted through the top electrode (TE) 6340.
  • Implant energies are sufficient to penetrate TE 6340 and into the switching layer (SL) or layers 6330 and into the bottom electrode (BE) 6320.
  • Ions can be inert such as Ar, He, Xe, for example.
  • transition metal switching layer(s) such as TiOx, HfOx, TaOx, ZrOx, WOx
  • O ions or transition metal ions can also be used such as Ti, Ni, Hf, W, Zr, Ta.
  • H and N can also be used to generate defects, vacancies, or ions to provide a path for electrical conduction along filaments.
  • Electrons hopping, tunneling, space charge, along the defects/vacancies provides a low resistance state (LRS) and when the vacancies have a gap, provide a high resistance state (HRS) of the RRAM.
  • LRS low resistance state
  • HRS high resistance state
  • ions are implanted through the top electrode, some of the top electrode ions are knocked into the switching layer, forming a more conductive layer at the TE-SL interface.
  • the TE is removed or partially removed typically by etching.
  • a nonlinear layer or layers 6360 is deposited on the SL 6330 to provide nonlinearity in the current voltage characteristics of the RRAM which is useful to minimize sneak current in crossbar RRAM architecture or in any array of RRAM used for memory such as vertical RRAM (VRRAM).
  • the nonlinear layer 6360 can be a diode layer of p and n doping for example in amorphous Si, or a layer of semiconductor material where Schottky barrier junctions can be created with contact to metal or almost metal surfaces.
  • the nonlinear layer 6330 can be a tunnel layer, can consist of multiple layers and/or combinations of diode and tunnel layers with Schottky junctions, metal oxide semiconductor junctions and/or tunnel junctions.
  • the TE 6342 is re-deposited on the diode layer 6360 to complete the RRAM cell.
  • the role of knocked on metal ions into the SL interface is important to allow a metal-like contact or a layer of higher conductivity next to the diode layer. This could be important, for example, in creating a metal-like interface between the SL and a semiconductor junction such as a Schottky barrier junction and/or when the diode layer is doped with p and n dopants.
  • the metal-like interface at the SL can provide a low resistance contact for example.
  • the TE can be thin (on the order of 2-50 nm for example) or thicker TE of 100-200 nm.
  • the metal-like interface formed at the TE-SL interface can then provide a good semiconductor contact with subsequent non-linear layers such as diode layer.
  • the TE 6340 can be on the order of one nm thick, in which case, after ion implant the TE does not need to be removed in the areas in contact with the SL.
  • the diode layer and/or other non-linear layers can be deposited onto the SL forming Schottly barrier junctions and/or almost ohmic or ohmic junctions depending on the doping of the diode layer or semiconductor layer. Note that the diode layer is a semiconductor layer but with p and n doping, whereas a semiconductor layer can be just doped p or n).
  • FIG. 64A-64B are diagrams illustrating ions knocked into an SL layer by ion implantation through a TE layer, according to some embodiments.
  • FIG. 64A shows the knocked in ions 6442 in the interface of SL layer 6430 after the TE has been removed.
  • FIG. 64B shows a residual TE 6444 remaining, either by partial removal and or by having a very thin TE prior to implant.
  • the TE layer can be approximately one nanometer, and the TE over the SL is not removed after ion implant.
  • the knocked in ions 6442 are from the TE due to ion
  • the knocked in ions will be W ions in the TE-SL interface with some W ions in the SL layer.
  • ion implanted through the TE for example Ar ions implanted through W TE physically knock the W ions into the SL in a ballistic fashion.
  • the Ar ions implanted into the SL 6430 and BE 6420 generate defects and vacancies that are conducive for filament formation, for species of ions such as oxygen, vacancies, and electrons to diffuse, drift, or hop in an electric field.
  • the diffusion, drifting and/or hopping can occur with joule heating due to current flow since power dissipated is proportional to the square of the current times the resistance.
  • FIGs. 65A-65B are diagrams illustrating ions knocked into an SL layer with a diode, semiconductor, and/or tunnel layer deposited, according to some embodiments.
  • FIG. 65A shows most of the TE removed and with the diode layer 6460 in contact with the SL 6430.
  • a residual TE 6444 is left, approximately one nm, and the DL 6460 is deposited on the residual TE and knocked in metal ions 6442 in the SL 6430.
  • additional metal ions and or oxygen ions can be implanted through the residual TE 6444 prior to deposited the diode or semiconductor layer 6460.
  • FIGs. 66A-66B are diagrams illustrating ions knocked into an SL layer with a diode, semiconductor, and/or tunnel layer, and TE deposited, according to some embodiments.
  • FIG. 66A shows the knocked in ions 6442 in the interface of SL layer 6430 after the TE has been removed, while FIG. 66B shows a residual TE 6444 remaining.
  • the top electrode 6462 is deposited on the diode and/or semiconductor layer 6460.
  • the diode and/or semiconductor layer 6460 can give a non-linear current voltage characteristics for the RRAM which is beneficial in reducing sneak current between adjacent RRAM cells in a high density array architecture, such as crossbar and 3D VRRAM stacks of RRAM cells.
  • the metal TE 6462 can form a Schottky junction with the diode/semiconductor layer 6460.
  • the metal-like interface between the SL 6430 and DL 6460 can have knocked in metal ions and the residual TE metal can also form a Schottky contact, MOS contact and/or ohmic contact, depending on the doping level of the diode/semiconductor layer.
  • FIG. 67 is a simple flow chart of the process of implanting through a TE, according to some embodiments.
  • the BE is deposited on the substrate.
  • the SL(s) are deposited.
  • a first TE is deposited.
  • ions are implanted through the TE.
  • the TE can either be left in place or it can be removed wholly or in part.
  • the diode/semiconductor layers are deposited.
  • the RRAM cell is completed with the deposition (or re- deposition) of the TE.
  • the metal-like interface between the SL and the DL separates the two functions, the SL providing creation and annihilation of vacancies to generate HRS and LRS and the diode/semiconductor layer(s) for giving a nonlinear current voltage characteristic of the RRAM cell.
  • the knocked-in ion techniques described can be applied to vertical RRAMs.
  • the vertical electrode 5320 can initially be a thin conformal layer such as W, TiN, Mo, Al, Ni to name a few that can be deposited by sputtering, atomic layer deposition for example. Ion implantation is then performed at a angle off normal with the wafer rotating for example, that can knock in metal ions into the interface of the SL 5330.
  • very thin initial vertical electrode 5320 for example approximately one nm, the electrode can be left in place while a
  • diode/semiconductor layer are deposited conformal to the SL 5330 and followed by a full version of the electrode 5320.

Landscapes

  • Engineering & Computer Science (AREA)
  • Manufacturing & Machinery (AREA)
  • Semiconductor Memories (AREA)

Abstract

La présente invention concerne des dispositifs de RAM résistive (RRAM) ayant une uniformité augmentée et des procédés de fabrication associés. Une plus grande uniformité de performances sur une puce entière qui comprend des nombres plus élevés de cellules RRAM peut être obtenue par création uniforme de canaux augmentés dans les couches de commutation au moyen de l'utilisation de dommages causés par un rayonnement. Le rayonnement, selon différents modes de réalisation décrits, peut être sous la forme d'ions, de photons électromagnétique, de particules neutres, d'électrons et d'ultrasons.
PCT/US2015/045731 2015-01-05 2015-08-18 Mémoire vive résistive avec canaux implantés et rayonnés WO2016111724A1 (fr)

Priority Applications (3)

Application Number Priority Date Filing Date Title
CN201580027848.6A CN106463340A (zh) 2015-01-05 2015-08-18 具有注入及辐射通道的电阻式随机存取存储器
JP2016569376A JP2018516447A (ja) 2015-01-05 2015-08-18 注入型及び照射型チャンネルを具備する抵抗ランダムアクセスメモリ
EP15877278.0A EP3243211B1 (fr) 2015-01-05 2015-08-18 Mémoire vive résistive avec des canaux implantés

Applications Claiming Priority (16)

Application Number Priority Date Filing Date Title
US201562100028P 2015-01-05 2015-01-05
US62/100,028 2015-01-05
US62/132,507 2015-02-04
US201562132507P 2015-03-13 2015-03-13
US201562137282P 2015-03-24 2015-03-24
US62/137,282 2015-03-24
US201562144328P 2015-04-07 2015-04-07
US62/144,328 2015-04-07
US201562145450P 2015-04-09 2015-04-09
US62/145,450 2015-04-09
US201562151394P 2015-04-22 2015-04-22
US62/151,394 2015-04-22
US201562171209P 2015-06-04 2015-06-04
US62/171,209 2015-06-04
US201562172110P 2015-06-07 2015-06-07
US62/172,110 2015-06-07

Publications (1)

Publication Number Publication Date
WO2016111724A1 true WO2016111724A1 (fr) 2016-07-14

Family

ID=56358354

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2015/045731 WO2016111724A1 (fr) 2015-01-05 2015-08-18 Mémoire vive résistive avec canaux implantés et rayonnés

Country Status (3)

Country Link
JP (1) JP2018516447A (fr)
CN (1) CN106463340A (fr)
WO (1) WO2016111724A1 (fr)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109411602A (zh) * 2018-11-22 2019-03-01 上海华力微电子有限公司 氧化钽基阻变存储器及其制造方法
WO2019146268A1 (fr) * 2018-01-25 2019-08-01 ソニーセミコンダクタソリューションズ株式会社 Élément de stockage et dispositif de stockage
US10381557B2 (en) 2015-12-14 2019-08-13 Shih-Yuan Wang Resistive random-access memory with protected switching layer
CN111403599A (zh) * 2020-02-26 2020-07-10 浙江省北大信息技术高等研究院 一种半导体结构及其制备方法
US10756266B2 (en) 2015-12-14 2020-08-25 Wwram Devices, Inc. Resistive random-access memory with protected switching layer
WO2020228889A1 (fr) * 2019-05-16 2020-11-19 Rheinisch-Westfälische Technische Hochschule (Rwth) Aachen Mémoire à changement de phase
US11152566B2 (en) 2019-12-10 2021-10-19 Winbond Electronics Corp. Resistive random access memory
US11258011B2 (en) 2019-07-01 2022-02-22 Winbond Electronics Corp. Resistive random access memory structure and manufacturing method thereof

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020024221A1 (fr) 2018-08-02 2020-02-06 深圳市为通博科技有限责任公司 Procédé de préparation et dispositif de préparation de matériau d'électrode de memristance et matériau d'électrode de memristance
CN109545964B (zh) * 2018-12-14 2021-03-19 中国科学院上海微系统与信息技术研究所 一种基于氧化物离子注入的选通材料、选通器单元及其制备方法
CN110676374B (zh) * 2019-05-22 2020-06-09 集美大学 一种阻变存储器及其制备方法
CN110783453B (zh) * 2019-09-24 2024-02-09 北京大学 一种双模阻变存储器件及其制备方法
CN111009609B (zh) * 2019-12-24 2022-06-07 华中科技大学 一种超晶格忆阻器功能层材料、忆阻器单元及其制备方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090272961A1 (en) * 2008-05-01 2009-11-05 Michael Miller Surface treatment to improve resistive-switching characteristics
US20100243983A1 (en) * 2009-03-31 2010-09-30 Tony Chiang Controlled localized defect paths for resistive memories
US20120252194A1 (en) * 2011-03-28 2012-10-04 Sen Corporation Ion implantation method and ion implantation apparatus
US20120295398A1 (en) 2011-05-16 2012-11-22 Varian Semiconductor Equipment Associates, Inc. Ion implant modification of resistive random access memory devices
US20130292632A1 (en) * 2009-04-10 2013-11-07 Intermolecular, Inc. Resistive Switching Memory Element Including Doped Silicon Electrode

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20080064353A (ko) * 2007-01-04 2008-07-09 삼성전자주식회사 저항 메모리 소자 및 그 제조 방법
CN102254803B (zh) * 2011-08-04 2013-05-01 江苏畅微电子科技有限公司 电阻型存储器的制备方法
WO2014194069A2 (fr) * 2013-05-29 2014-12-04 Shih-Yuan Wang Mémoire à accès aléatoire résistive formée sans tension d'écriture

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090272961A1 (en) * 2008-05-01 2009-11-05 Michael Miller Surface treatment to improve resistive-switching characteristics
US20100243983A1 (en) * 2009-03-31 2010-09-30 Tony Chiang Controlled localized defect paths for resistive memories
US20130292632A1 (en) * 2009-04-10 2013-11-07 Intermolecular, Inc. Resistive Switching Memory Element Including Doped Silicon Electrode
US20120252194A1 (en) * 2011-03-28 2012-10-04 Sen Corporation Ion implantation method and ion implantation apparatus
US20120295398A1 (en) 2011-05-16 2012-11-22 Varian Semiconductor Equipment Associates, Inc. Ion implant modification of resistive random access memory devices

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10381557B2 (en) 2015-12-14 2019-08-13 Shih-Yuan Wang Resistive random-access memory with protected switching layer
US10756266B2 (en) 2015-12-14 2020-08-25 Wwram Devices, Inc. Resistive random-access memory with protected switching layer
KR20200110330A (ko) * 2018-01-25 2020-09-23 소니 세미컨덕터 솔루션즈 가부시키가이샤 기억 소자 및 기억 장치
JP2019129239A (ja) * 2018-01-25 2019-08-01 ソニーセミコンダクタソリューションズ株式会社 記憶素子および記憶装置
WO2019146268A1 (fr) * 2018-01-25 2019-08-01 ソニーセミコンダクタソリューションズ株式会社 Élément de stockage et dispositif de stockage
US11522132B2 (en) 2018-01-25 2022-12-06 Sony Semiconductor Solutions Corporation Storage device and storage unit with a chalcogen element
KR102631895B1 (ko) 2018-01-25 2024-02-01 소니 세미컨덕터 솔루션즈 가부시키가이샤 기억 소자 및 기억 장치
CN109411602A (zh) * 2018-11-22 2019-03-01 上海华力微电子有限公司 氧化钽基阻变存储器及其制造方法
WO2020228889A1 (fr) * 2019-05-16 2020-11-19 Rheinisch-Westfälische Technische Hochschule (Rwth) Aachen Mémoire à changement de phase
US11817146B2 (en) 2019-05-16 2023-11-14 Rheinisch-Westfälische Technische Hochschule (Rwth) Aachen Phase-change memory
US11258011B2 (en) 2019-07-01 2022-02-22 Winbond Electronics Corp. Resistive random access memory structure and manufacturing method thereof
US11152566B2 (en) 2019-12-10 2021-10-19 Winbond Electronics Corp. Resistive random access memory
CN111403599A (zh) * 2020-02-26 2020-07-10 浙江省北大信息技术高等研究院 一种半导体结构及其制备方法
CN111403599B (zh) * 2020-02-26 2022-11-04 杭州未名信科科技有限公司 一种半导体结构及其制备方法

Also Published As

Publication number Publication date
JP2018516447A (ja) 2018-06-21
CN106463340A (zh) 2017-02-22

Similar Documents

Publication Publication Date Title
US9647036B2 (en) Resistive random-access memory with implanted and radiated channels
US9978939B2 (en) Resistive random-access memory with implanted and radiated channels
WO2016111724A1 (fr) Mémoire vive résistive avec canaux implantés et rayonnés
JP5507545B2 (ja) 抵抗スイッチング特性を改善する表面処理方法
KR100870054B1 (ko) 고체 전해질 메모리 소자 및 이러한 메모리 소자를제조하는 방법
US20120295398A1 (en) Ion implant modification of resistive random access memory devices
US8728839B2 (en) Memory cells and methods of forming memory cells
US9159769B2 (en) Semiconductor device, semiconductor device manufacturing method, and semiconductor device manufacturing apparatus
US10381557B2 (en) Resistive random-access memory with protected switching layer
EP2827395A1 (fr) Procédé de modelage d'un empilement de jonctions à effet tunnel magnétique
TW200828637A (en) Method for making a self-converged memory material element for memory cell
US8501621B2 (en) Method of fabrication of the memristive device
US8502188B2 (en) Electrically actuated device
US8809159B2 (en) Radiation enhanced resistive switching layers
US7718537B2 (en) Method for manufacturing a CBRAM semiconductor memory
EP3243211B1 (fr) Mémoire vive résistive avec des canaux implantés
Kim et al. Ion beam-assisted solid phase epitaxy of SiGe and its application for analog memristors
Taggart et al. Resistance state locking in CBRAM cells due to displacement damage effects
US8461565B2 (en) Electrically actuated devices
Pattnaik et al. Effect of gamma radiation on electrical properties of diffusive memristor devices
JP5614810B2 (ja) 注入方法
CN115224193A (zh) 一种忆阻器及其制备方法
KR20220064728A (ko) 다중층 선택 소자 및 그 제조 방법
TW201126522A (en) Memory device with field enhancement arrangement

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15877278

Country of ref document: EP

Kind code of ref document: A1

REEP Request for entry into the european phase

Ref document number: 2015877278

Country of ref document: EP

WWE Wipo information: entry into national phase

Ref document number: 2015877278

Country of ref document: EP

ENP Entry into the national phase

Ref document number: 2016569376

Country of ref document: JP

Kind code of ref document: A

NENP Non-entry into the national phase

Ref country code: DE